diff --git "a/2dE4T4oBgHgl3EQfagyV/content/tmp_files/load_file.txt" "b/2dE4T4oBgHgl3EQfagyV/content/tmp_files/load_file.txt"
new file mode 100644--- /dev/null
+++ "b/2dE4T4oBgHgl3EQfagyV/content/tmp_files/load_file.txt"
@@ -0,0 +1,2701 @@
+filepath=/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf,len=2700
+page_content='Toward Building General Foundation Models for Language, Vision, and  Vision-Language Understanding Tasks  Xinsong Zhang 1 Yan Zeng 1 Jipeng Zhang 2 Hang Li 1  Abstract  Foundation models or pre-trained models have  substantially improved the performance of various  language, vision, and vision-language understand-  ing tasks.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' However, existing foundation models  can only perform the best in one type of tasks,  namely language, vision, or vision-language.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' It is  still an open question whether it is possible to con-  struct a foundation model performing the best for  all the understanding tasks, which we call a gen-  eral foundation model.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' In this paper, we propose  a new general foundation model, X-FM (the X-  Foundation Model).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' X-FM has one language en-  coder, one vision encoder, and one fusion encoder,  as well as a new training method.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' The training  method includes two new techniques for learning  X-FM from text, image, and image-text pair data.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='  One is to stop gradients from the vision-language  training when learning the language encoder.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' The  other is to leverage the vision-language training  to guide the learning of the vision encoder.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' Exten-  sive experiments on benchmark datasets show that  X-FM can significantly outperform existing gen-  eral foundation models and perform better than or  comparable to existing foundation models specif-  ically for language, vision, or vision-language  understanding.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='  1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' Introduction  With the enormous power of foundation models, also known  as pre-trained models, remarkable performance gains have  recently been achieved in a variety of understanding tasks in  natural language processing (NLP), computer vision (CV),  and other fields (Devlin et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', 2019;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' Liu et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', 2019;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' Lewis  et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', 2020;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' Raffel et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', 2020;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' Brown et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', 2020;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' Doso-  vitskiy et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', 2021;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' He et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', 2022;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' Bao et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', 2021;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' Lu  1ByteDance AI Lab 2The Hong Kong University of Science  and Technology.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' Correspondence to: Xinsong Zhang <zhangxin-  song.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='0320@bytedance.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='com>.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='  Copyright 2023 by the author(s).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' The code and pre-trained models  will be released upon publication.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='  et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', 2019;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' Tan & Bansal, 2019a;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' Chen et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', 2020;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' Li  et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', 2020;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' 2021a;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' Zeng et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', 2021;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' 2022) .' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' Foundation  models are usually equipped with Transformer (Vaswani  et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', 2017) as the backbone, pre-trained with a tremendous  amount of unlabeled data, and then fine-tuned with small  amounts of labeled data in downstream tasks.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' The strong  representation ability of the model, the massive amount of  data, and the effective means of training make the founda-  tion models powerful for successfully solving the tasks of  vision, language, and vision-language (Li et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', 2021b;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='c;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='  Singh et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', 2021;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' Wang et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', 2021b;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' 2022b;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' Diao et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=',  2022;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' Wang et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', 2022a).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='  The state-of-the-art foundation models usually work the  best for one type of tasks, namely language, vision, and  vision-language.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' For example, RoBERTa (Liu et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', 2019),  BEiTv2 (Peng et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', 2022), and X-VLM (Zeng et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', 2021;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='  2022) are language, vision, and vision-language founda-  tion models respectively, and can achieve state-of-the-art  performances for the specific type of tasks.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' It is still very  challenging, however, to build a general foundation model  that can perform the best in all types of tasks.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' Existing  models, such as FLAVA (Singh et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', 2021), OFA (Wang  et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', 2022b), DaVinci (Diao et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', 2022) and Uni-Perceiver-  MoE (Zhu et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', 2022), are trying to achieve the goal.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' Their  performances are still not satisfactory, however, when com-  pared with the best performing foundation models for the  individual types of tasks, as shown in Table 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' Previous  work (Bingel & Søgaard, 2017;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' Wang et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', 2020) also  shows that it is difficult to train a general foundation model  in a multi-task learning setting that can effectively learn and  utilize representations for all types of tasks.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' The reason is  that language, vision, and vision-language are very different  in nature, and a simple way of jointly training a model from  language, vision, and vision-language data can easily create  a suboptimal solution.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='  To address the challenge, we propose a new general founda-  tion model, X-FM (X-Foundation Model).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' X-FM consists of  three modular encoders for language (text) encoding, vision  (image) encoding, and fusion encoding, as shown in Fig 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='  The language encoder, the vision encoder, and the entire  model can be used in downstream tasks of language, vision,  and vision-language understanding, respectively.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' All three  arXiv:2301.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='05065v1  [cs.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='CV]  12 Jan 2023  Toward Building General Foundation Models for Language, Vision, and Vision-Language Understanding Tasks  Methods  Text Tasks  Vision Tasks  Multi-modal Tasks (MSCOCO Retriveal & VQA)  GLUE  ImageNet  Zero-Shot  Fine-Tune  MNLI  RTE  FT/LE  TR  IR  TR  IR  VQA  Foundation models specifically for language, vision, or vision-language understanding  RoBERTa (Liu et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', 2019)  87.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='6  78.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='7  –  –  –  –  –  –  BEiTv2 (Peng et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', 2022)  –  –  85.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='5/80.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='1  –  –  –  –  –  X-VLM (Zeng et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', 2021)  –  –  –  70.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='8/92.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='1/96.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='5  55.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='6/82.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='7/90.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='0  80.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='4/95.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='5/98.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='2  63.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='1/85.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='7/91.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='6  78.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='1  X2-VLM (Zeng et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', 2022)  –  –  –  –  –  80.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='5/95.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='5/97.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='8  62.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='7/84.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='7/90.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='7  79.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='2  General foundation models  UNIMO-2 (Li et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', 2021c)  87.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='5  –  80.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='8/-  –  –  –  –  76.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='3  SimVLM (Wang et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', 2021c)  83.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='4  63.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='9  /80.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='6  –  –  –  –  77.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='9  FLAVA (Singh et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', 2021)  80.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='3  57.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='8  /75.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='5  42.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='7/76.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='8/-  38.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='4/67.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='5/-  61.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='5/82.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='1/89.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='6  50.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='1/74.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='4/83.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='2  72.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='8  OFA (Wang et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', 2022b)  84.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='3  70.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='8  82.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='2/–  –  –  –  –  78.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='0  DaVinci (Diao et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', 2022)  83.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='1  64.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='2  83.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='9/78.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='8  –  –  –  –  76.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='3  OmniVL (Wang et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', 2022a)  –  –  –  –  –  76.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='8/93.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='6/97.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='3  58.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='5/82.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='6/89.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='5  78.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='3  Uni-Perceiver-MoE (Zhu et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', 2022)  81.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='5  75.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='8  84.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='5/–  64.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='6/–/–  51.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='6/–/–  70.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='5/–/–  54.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='1/–/–  –  X-FMbase  87.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='7  83.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='2  85.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='3/81.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='0  73.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='8/93.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='9/97.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='2  59.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='4/83.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='6/90.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='0  81.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='8/96.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='0/98.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='3  64.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='7/86.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='1/91.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='6  79.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='1  Table 1: Performance comparisons between foundation models.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' All results are from base-size models.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' MSCOCO is a  cross-modal retrieval task, and IR and TR are image-retrieval and text-retrieval, respectively.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' MNLI results are average  accuracies of MNLI-m and MNLI-mm.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' Accuracy is reported for RTE.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' For ImageNet1k classification, we report linear  evaluation (LE) performance and fine-tuning (FT) performance, respectively.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' We report R@1/R@5/R@10 for all retrieval  tasks at both zero-shot and fine-tune settings.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' We report the VQA test-dev result.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' bold denotes the best number across  general foundation models.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' underline denotes the best across all models.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='  encoders are stacked Transformer layers.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' The language en-  coder and the vision encoder follow the implementations  of BERT (Devlin et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', 2019) and ViT (Dosovitskiy et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=',  2021), respectively.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' The fusion encoder has the same ar-  chitecture as BERT except that there is a cross-attention  sub-layer after the self-attention sub-layer in each Trans-  former layer.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='  In learning of X-FM, the language encoder, vision encoder,  and fusion encoder are jointly trained with text data, im-  age data, and image-text pair data as input.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' Given the text  data, we train the language encoder by masked language  modeling (MLM).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' Given the image data, we train the vi-  sion encoder by masked image modeling (MIM).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' Given the  image-text pair data, we train the fusion encoder by image  text matching (ITM), image-conditioned masked language  modeling (IMLM), bounding box prediction (BBP), train  the vision encoder and the language encoder by image-text  contrastive learning (ITC), and train the vision encoder by  MIM.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' (See Fig 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=')  The essential thinking of our learning method is that lan-  guage is more abstract than vision, and there is an asymmet-  ric relationship between language and vision.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' Therefore, we  separate the learning of the three encoders.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' The language  encoder is trained mainly from text data and is isolated from  the training of the fusion encoder.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' The vision encoder is  simultaneously trained from image data and image-text pair  data, guided by the vision-language training.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' The fusion  encoder is trained from image-text pair data.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='  Our learning method includes two new techniques.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' One  technique is to stop gradients from the vision-language train-  ing when learning the language encoder.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' The gradient flow  is stopped from the fusion encoder to the language encoder  in training, while the activation flow from the language en-  coder to the fusion encoder is as usual.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' As a result, the  language encoder is not affected by training of the fusion  encoder with image-text pair data.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' Moreover, the training of  the fusion encoder concentrates on learning the alignments  between language and vision features.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='  The other technique is to leverage the vision-language train-  ing to guide the learning of the vision encoder with masked  image modeling (MIM).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' In MIM, the masked image is com-  pared with the original image by the differences between the  predicted representations and target representations at the  masked and [CLS] positions.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' The vision encoder creates  both the predicated and target representations, while there  is gradient flow from the predicted representations but no  gradient flow from the target representations.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' The vision  encoder can create the target representations because it is  also trained in the vision-language training.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='  We conduct experiments on a variety of twenty-two tasks of  language, vision, and vision-language understanding.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' X-FM  can outperform other general foundation models by a large  margin and can even achieve better or comparable perfor-  mance than SOTA foundation models specifically designed  for language, vision, or vision-language understanding tasks,  as shown in Table 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='  Toward Building General Foundation Models for Language, Vision, and Vision-Language Understanding Tasks  2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' Related Work  Following the success of language model pre-training, vi-  sion pre-training and vision-language pre-training with  Transformer as the backbone (Vaswani et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', 2017) have  also made significant progress recently, pushing the state-of-  the-art of various understanding tasks of language, vision,  and vision-language.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='  In language understanding, BERT (Devlin et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', 2019) is  the first model adopting masked language modeling (MLM)  for pre-training, which achieves remarkable performance  on a wide range of tasks.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' Several other models are then  developed to improve training robustness (Liu et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', 2019),  sample efficiency (Sun et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', 2019;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' Joshi et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', 2020;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' Clark  et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', 2020), and prediction accuracy of BERT (Lan et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=',  2020;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' Zhang et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', 2020;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' He et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', 2021).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='  In vision understanding, ViT (Dosovitskiy et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', 2021;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' Tou-  vron et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', 2021) is proposed, utilizing Transformer as the  backbone.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' Inspired by MLM, subsequent work proposes  using masked image modeling (MIM) with the objective of  recovering masked images.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' The learning targets vary from  pixels (He et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', 2022) to image tokens (Bao et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', 2021;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='  Peng et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', 2022).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='  In vision-language understanding, there are generally two  approaches.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' One is “dual encoders,” in which image and text  are encoded separately, followed by a shallow interaction  layer.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' The other is “fusion encoder(s)” in which attention  or self-attention is used to fuse information from the two  modalities after encoding.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' The former approach includes  CLIP (Radford et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', 2021) and ALIGN (Jia et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', 2021)  and performs well in vision tasks and cross-modal retrieval  tasks.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' However, it cannot perform so well in multi-modal  fusion tasks such as visual question answering (VQA (Goyal  et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', 2017)) and visual reasoning (NLVR2 (Suhr et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=',  2019b)).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' The latter approach varies depending on the way  of using image features.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' Early work feeds pre-extracted  object features along with texts into Transformer models  and trains the models to make multi-modal modeling and  multi-modal alignments with suitable objectives (Lu et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=',  2019;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' Tan & Bansal, 2019b;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' Li et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', 2020;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' Chen et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=',  2020;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' Cho et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', 2021;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' Zhang et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', 2021).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' Later work  uses patch embeddings directly with new architectures such  as vision Transformer (Li et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', 2021a;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' 2022) or multiway  Transformer (Wang et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', 2021a;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' Bao et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', 2022) and uses  new objectives such as bounding box prediction (Zeng et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=',  2021;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' 2022).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='  Recently, the fact that Transformer can model multi-modal  data within a single architecture has inspired research to  develop general foundation models that can solve lan-  guage, vision, and vision-language tasks at the same time.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='  UNIMO (Li et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', 2021b;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='c) jointly learns from image  and text data vision representations, language representa-  tions, and vision-language alignments in a shared space.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='  FLAVA (Singh et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', 2021), a general foundation model, per-  forms pre-training with masked uni-modal and multi-modal  modeling objectives.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' OFA (Wang et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', 2022c) formulates  vision-language tasks as sequence-to-sequence (seq2seq)  problems and pre-trains a seq2seq model in multi-task learn-  ing.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' SimVLM (Wang et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', 2021c) pre-trains a seq2seq  model with a single objective of language generation (prefix  language modeling).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' DaVinci (Diao et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', 2022) combines  prefix language modeling and prefix image modeling to  learn a general foundation model for a wide range of tasks.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='  Uni-Perceiver (Zhu et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', 2021;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' 2022) builds a unified per-  ception architecture that processes various modalities and  tasks with a single Transformer network and shared parame-  ters.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='  Previous studies on general foundation models have shown  that different capabilities can be established with only one  model.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' Still, few studies demonstrate that the best perfor-  mance can be achieved in all tasks with one model.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' In this  paper, we propose a new general foundation model and show  that it can perform the best for all the understanding tasks  of language, vision, and vision-language.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' We compare our  model extensively with recent general foundation models  on multiple dimensions, as shown in Appendix A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='  Several super-large foundation models (over 1B parame-  ters) are proposed recently, most of which are trained on  super-large in-house datasets (over 400M image-text pairs).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='  The authors do not report results at the base (about 280M  parameters) and large (about 800M parameters) scale on  public datasets, which we consider in this paper.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' CoCa (Yu  et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', 2022) pre-trains an image-text sequence-to-sequence  model with contrastive loss and captioning loss.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' BEiT-  3 (Wang et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', 2022d) uses a multi-way Transformer and a  unified objective of masked “language” modeling for learn-  ing from image (Imglish1), text, and image-text pair data.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='  Florence (Yuan et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', 2021) first scales the web-scale image-  text pairs to 900M representations and then adapts to vari-  ous computer vision tasks.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' Flamingo (Alayrac et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', 2022)  makes use of a large language model in vision-language  pre-training to solve the “in-context learning” problem for  vision-language tasks.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' PaLI (Chen et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', 2022) jointly scales  up the vision encoder and language encoder to cover a vari-  ety of language, vision, vision-language, and multilingual  tasks.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='  3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' Method  3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' Model Architecture and Training Process  We propose a new general foundation model X-FM, having  a language encoder, a vision encoder, and a fusion encoder,  shown as Fig 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' The language encoder is a stack of Trans-  1They view the image as a foreign language.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='  Toward Building General Foundation Models for Language, Vision, and Vision-Language Understanding Tasks  Text Encoder  Cross-Attention  Fusion Encoder  Feed Forward  Self-Attention  Vision Encoder  Feed Forward  Self-Attention  two  brown  and  white  dogs.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='  Mask  Feed Forward  Self-Attention  stop grad  MIM  MLM  IMLM, ITM, BBP  ITC  x M  x N  x L  Mask  target  Figure 1: The architecture and pre-training process of X-FM, a Transformer-based general foundation model.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' Given  a text, we learn the language encoder by MLM.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' Given an image, we learn the vision encoder by MIM.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' Given an image-text  pair, we learn the fusion encoder by BBP, ITM, IMLM and ITC, and further learn the vision encoder by MIM.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' The gradients  of BBP, ITM, and IMLM are stopped from the fusion encoder to the language encoder.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' The vision encoder is trained by  MIM with both the image-text pair data and the image data.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' M, N and P denote numbers of encoder layers.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='  former layers like that of BERT (Devlin et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', 2019), while  the vision encoder is a stack of Transformer layers like that  of ViT (Dosovitskiy et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', 2021).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' The language encoder uses  a post-layer-norm, while the vision encoder uses a pre-layer-  norm.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' The fusion encoder is similar to that of ALBEF (Li  et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', 2021a) and X-VLM (Zeng et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', 2021), in which  each layer has an attention sub-layer after a self-attention  sub-layer.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' In the self-attention sub-layers, the queries are  from language and the keys & values are from vision.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='  We propose a new method for learning X-FM, also shown in  Fig 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' Text, image, and image-text pair data are used as input  to train X-FM.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' The language encoder is trained by masked  language modeling (MLM) and image text contrastive learn-  ing (ITC).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' The vision encoder is trained by masked image  modeling (MIM) and ITC.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' The fusion encoder is trained  by image text matching (ITM), image-conditioned masked  language modeling (IMLM), and bounding box prediction  (BBP).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' There are two new techniques developed for the  training.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='  Stop Gradient.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' We stop gradients from the vision-language  training when learning the language encoder.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' Specifically,  when the fusion encoder is trained with image-text pair  data by ITM, IMLM, and BBP, there are forward flows  (activations) from the language encoder to the fusion en-  coder, but there are no backward flows (gradients) from the  fusion encoder to the language encoder.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' In this way, the  language encoder is only trained with text data by MLM  and with image-text pair data by ITC.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' The former helps the  language encoder to learn text representations, and the latter  helps the language encoder and the vision encoder to make  alignments between their respective text representations and  image representations.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' Meanwhile, the training of the fusion  encoder is performed separately with the focus of learning  from image-text pair data.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='  Masked Image Modeling.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' The training of vision encoder  by MIM is carried out as follows.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' The image data is first  masked and then predicted by the vision encoder.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' The dif-  ferences between predicted representations and ‘target’ rep-  resentations at masked positions and [CLS] position are  then measured with MSE (mean squared error) loss.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' The  target representations are obtained from the same image  data (without masking) by the vision encoder.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' There are  no gradients from the target representations in the learning  of the vision encoder.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' The vision encoder can create target  representations because it is also trained with image-text  pair data.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' In this way, the vision encoder is trained by both  the cross-modal objectives (ITC, ITM, BBP, IMLM) with  image-text pair data and the uni-modal objective (MIM)  with image data.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' The representations obtained from the  vision-language training are highly semantic, which is nec-  essary for MIM as demonstrated in previous work (Bao  et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', 2021;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' Peng et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', 2022;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' Wei et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', 2022a;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='b).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='  There are three advantages by exploiting the new MIM  technique.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' First, it becomes possible to leverage image data  for learning of the vision encoder, which is relatively easy  to obtain.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' Second, it is convenient to conduct MIM with the  signals from the vision-language training.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' Note that most  previous work for MIM makes use of an external image  tokenizer such as VQ-VAE (Bao et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', 2021;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' Singh et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=',  2021), CLIP (Wei et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', 2022b), and VQ-KL (Peng et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=',  Toward Building General Foundation Models for Language, Vision, and Vision-Language Understanding Tasks  2022).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' Third, the learning of the vision encoder and that  of the fusion encoder are mutually enhanced.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' Once the  vision encoder is trained, it is also utilized to train the fusion  encoder.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='  3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' Pre-training Objectives  We explain six objectives in learning of X-FM.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' Here, T  represents the distribution of text data, I represents the  distribution of image data, and D represents the distribution  of image-text pair data.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='  Masked Language Modeling (MLM) We perform MLM  on text data to learn the language encoder of X-FM.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' Specifi-  cally we recover the masked tokens in a text by minimizing  the cross entropy loss below.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='  Lmlm = ET ∼T H(⃗y( ¯T), ˆ⃗p( ¯T))  (1)  where T denotes a text, ¯T denotes the masked text of T, ˆ⃗p  denotes the predicted probability vectors of masked tokens  of ¯T, ⃗y denotes the one-hot vectors representing the original  tokens of ¯T, and H denotes cross-entropy.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='  Image-Text Contrastive Learning (ITC).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' We use an  image-text contrastive loss as in CLIP (Radford et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', 2021)  to learn the alignments between images and texts in ITC.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='  Given a batch of images and texts, we calculate the cosine  similarities between all image-text pairs.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' For each image,  there is one text matched and the rest is unmatched.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' For each  text, there is one image matched and the rest is unmatched.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='  The contrastive loss is defined as follows.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='  Litc = 1  2E(I,T )∼D  �  H(⃗yi2t(I), ⃗pi2t(I))  + H(⃗yt2i(T), ⃗pt2i(T))  �  (2)  where (I, T) denotes an image-text pair, ⃗pi2t(I) denotes the  in-batch image-to-text similarities, ⃗pt2i(T) denotes the in-  batch text-to-image similarities, ⃗yi2t(I) denotes the one-hot  vectors representing the image-to-text matching relations,  ⃗yt2i(T) denotes the one-hot vectors representing the text-to-  image matching relations, and H denotes cross-entropy.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='  Image-Text Matching (ITM).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' We also learn the align-  ments between images and texts in ITM, using a loss in-  dicating whether an image-text pair is matched.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' For each  image in a batch there is a matched (positive) text, and we  sample an unmatched (negative) text in the batch.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' For each  text there is a matched (positive) image, and we sample  an unmatched image in the batch.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' The loss is defined as  follows.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='  Litm = E(I,T )∼D  �  H(pmatch(I, T))  +H(pmatch(˜I, T))  (3)  +H(pmatch(I, ˜T))  �  where (I, T) denotes a positive image-text pair, (˜I, T) and  (I, ˜T) denote negative image-text pairs, pmatch(I, T) de-  notes a predicted matching probability of (I, T), and H  denotes logistic loss.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='  Image-conditioned  Masked  Language  Modeling  (IMLM) We conduct IMLM on image-text pair data to  learn the fusion encoder.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='  Specifically, we recover the  masked tokens of the text given for an image-text pair by  minimizing the cross entropy loss below.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='  Limlm = E(I,T )∼DH(⃗y( ¯T), ˆ⃗p(I, ¯T))  (4)  where (I, T) denotes an image-text pair, ¯T denotes the  masked text of T, ˆ⃗p(I, ¯T) denotes the predicted probability  vectors of the masked tokens of ¯T based on I, ⃗y denotes the  one-hot vectors representing the original tokens of ¯T, and  H denotes cross-entropy.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='  Bounding Box Prediction (BBP) We adopt the BBP in X-  VLM (Zeng et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', 2021;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' 2022), which locates the visual  concept in the image by a bounding box given the text.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' With  BBP we learn the alignments between the images and texts  in multi-granularity.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' In BBP, two losses are simultaneously  minimized to measure the differences between the predicted  bounding box and the ground-truth bounding box.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' One  is generalized intersection over union GIoU (Rezatofighi  et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', 2019) and the other is ℓ1 distance.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='  Lbbp = E(I,T )∼D{GIoU(⃗b,ˆ⃗b) + ∥⃗b − ˆ⃗b∥1}  (5)  where⃗b = (cx, cy, w, h) denotes the ground truth bounding  box, ˆ⃗b = ( ˆcx, ˆcy, ˆw, ˆh) denotes the predicted bounding box.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='  A bounding box is represented by two coordinates, width,  and height.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='  Masked Image Modeling (MIM) We perform MIM on im-  age data and image-text pair data to learn the vision encoder.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='  Specifically, we recover the masked image patches in an  image by minimizing the loss below.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='  Lmim = E(I,T )∼D||⃗v(¯I) − ˆ⃗v(¯I)||2 + EI∼I||⃗v(¯I) − ˆ⃗v(¯I)||2  (6)  where (I, T) and I denote an image-text pair and a single  image respectively, ¯I denotes the masked image I, ˆ⃗v(¯I) de-  notes the predicted representations at the masked positions  and [CLS] of ¯I, and ⃗v(¯I) denotes the target representa-  tions at the masked positions and [CLS] of ¯I.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' ||˙||2 is the  MSE loss.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' We employ block masking following previous  work (Bao et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', 2021;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' Peng et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', 2022).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' Note that (I, T)  and I are independently sampled from D and I, and the  sample sizes are not necessarily equal.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='  Finally, the pre-training objective of X-FM is defined as the  sum of the losses described above.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='  L = Lmlm + Litc + Litm + Limlm + Lbbp + Lmim (7)  Toward Building General Foundation Models for Language, Vision, and Vision-Language Understanding Tasks  Base-Size Models  Large-Size Models  RoBERTa  BEiTv2  X2-VLM  UNIMO-2  FLAVA  SimVLM  OFA  DaVinci  Uni-Per.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='  OmniVL  X-FM  RoBERTa  BEiTv2  X2-VLM  SimVLM  OFA  Uni-Per.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='  X-FM  Task  Eval.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='  1  2  3  4  5  6  7  8  9  10  11  12  13  14  15  16  17  18  MNLI  FT  87.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='6  –  –  87.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='5  80.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='3  83.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='4  84.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='3  82.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='3  81.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='5  –  87.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='7  90.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='2  –  –  –  84.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='3  85.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='7  90.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='4  CoLA  FT  63.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='6  –  –  62.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='1  50.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='7  46.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='7  52.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='3  52.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='1  52.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='2  –  65.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='3  68.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='0  –  –  –  52.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='3  57.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='4  69.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='9  MRPC  FT  90.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='2  –  –  –  84.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='2  79.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='8  88.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='7  83.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='1  –  –  91.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='7  90.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='9  –  –  –  88.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='7  –  92.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='4  QQP  FT  91.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='9  –  –  –  88.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='7  90.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='4  91.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='3  88.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='2  –  –  91.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='8  92.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='2  –  –  –  91.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='3  –  92.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='2  SST-2  FT  94.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='8  –  –  94.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='7  90.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='9  90.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='9  92.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='7  90.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='5  90.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='9  –  95.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='0  96.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='4  –  –  –  92.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='7  93.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='4  96.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='7  QNLI  FT  92.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='8  –  –  –  87.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='3  88.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='6  91.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='1  87.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='2  88.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='2  –  92.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='9  94.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='7  –  –  –  91.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='1  91.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='9  94.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='8  RTE  FT  78.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='70  –  –  –  57.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='8  63.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='9  70.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='8  60.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='7  75.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='8  –  83.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='8  86.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='6  –  –  –  70.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='8  78.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='4  87.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='4  STS-B  FT  91.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='2  –  –  91.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='2  85.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='7  87.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='2  –  86.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='3  –  –  90.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='8  92.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='4  –  –  –  –  –  92.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='1  Language Avg.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='  86.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='4  –  –  –  78.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='2  78.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='9  –  78.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='8  –  –  87.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='4  88.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='9  –  –  –  –  –  89.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='5  ImageNet  FT  –  85.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='5  –  80.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='8  –  –  82.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='2  83.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='9  84.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='5  –  85.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='3  –  87.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='3  –  –  –  86.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='4  86.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='3  ImageNet  LE  –  80.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='1  –  –  75.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='5  80.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='6  71.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='4†  75.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='9  –  –  81.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='0  –  66.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='8†  –  82.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='3  74.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='7†  –  81.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='0  Food101  LE  –  88.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='2†  –  –  88.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='5  –  75.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='2†  89.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='3  –  87.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='4  88.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='7  –  52.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='2†  –  –  81.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='6†  –  88.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='9  CIFAR10  LE  –  95.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='3†  –  –  92.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='9  –  86.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='1†  93.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='0  –  96.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='2  97.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='2  –  63.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='5†  –  –  91.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='9†  –  97.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='2  CIFAR100  LE  –  81.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='5†  –  –  77.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='7  –  66.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='7†  79.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='0  –  83.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='2  86.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='7  –  39.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='7†  –  –  75.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='6†  –  85.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='1  Pets  LE  –  93.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='1†  –  –  84.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='8  –  81.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='0†  85.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='5  –  87.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='1  90.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='8  –  38.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='9†  –  –  86.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='8†  –  90.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='0  DTD  LE  –  78.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='4†  –  –  77.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='3  –  70.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='3†  77.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='1  –  76.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='2  78.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='4  –  44.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='4†  –  –  74.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='4†  –  79.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='0  Flowers102  LE  –  95.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='7†  –  –  96.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='4  –  86.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='3†  96.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='1  –  89.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='8  97.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='1  –  66.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='6†  –  –  92.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='6†  –  95.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='8  Vision Avg.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='  –  88.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='7  –  –  86.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='3  –  79.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='2  86.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='7  –  86.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='7  89.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='8  –  50.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='9  –  –  83.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='8  –  89.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='3  VQAv2  FT  –  –  79.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='2  76.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='3  72.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='5  77.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='9  78.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='0  73.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='9  –  78.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='3  79.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='1  –  –  80.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='5  79.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='3  80.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='3  –  79.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='5  NLVR2  FT  –  –  86.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='1  –  –  81.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='8  –  77.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='9  –  –  86.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='7  –  –  87.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='6  84.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='8  –  –  87.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='8  Flickr30K TR R@1  ZS  –  –  85.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='1†  88.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='5  67.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='7  –  –  –  82.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='1  –  90.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='1  –  –  86.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='8†  –  –  83.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='6  89.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='7  Flickr30K IR R@1  ZS  –  –  77.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='3†  72.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='7  65.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='2  –  –  –  72.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='4  –  79.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='1  –  –  80.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='5†  –  –  75.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='9  79.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='1  Flickr30K TR R@1  FT  –  –  97.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='4  92.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='0  –  –  –  –  93.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='6  94.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='9  97.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='4  –  –  99.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='1  –  –  94.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='1  97.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='9  Flickr30K IR R@1  FT  –  –  90.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='0  80.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='1  –  –  –  –  79.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='8  83.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='4  88.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='6  –  –  91.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='1  –  –  83.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='7  89.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='4  COCO TR R@1  ZS  –  –  68.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='4†  –  42.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='7  –  –  –  64.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='6  –  73.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='8  –  –  69.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='7†  –  –  67.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='9  74.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='4  COCO IR R@1  ZS  –  –  55.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='2†  –  38.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='4  –  –  –  51.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='6  –  59.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='4  –  –  58.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='3†  –  –  55.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='3  59.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='4  COCO TR R@1  FT  –  –  80.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='5  –  –  –  –  –  70.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='5  76.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='8  81.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='8  –  –  82.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='3  –  –  74.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='7  82.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='1  COCO IR R@1  FT  –  –  62.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='7  –  –  –  –  –  52.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='6  58.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='5  64.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='7  –  –  65.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='2  –  –  57.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='1  65.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='4  Vision-Language Avg.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='  –  –  78.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='2  –  –  –  –  –  –  –  80.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='1  –  –  80.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='1  –  –  –  80.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='5  Table 2: Experimental results on vision, language and vision-language tasks.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' MNLI results are average of MNLI-m  and MNLI-mm.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' MRPC results are average accuracies and F1 scores.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' Matthews correlation coefficient (MCC) is reported for  CoLA, and Pearson correlation coefficient (PCC) is reported for STS-B.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' We report accuracies for all the vision and multi-  modal tasks.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' FT is short for fine-tuning, LE for linear evaluation, ZS for zero-shot, TR for text retrieval, and IR for image  retrieval.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' Results for RoBERTa are from its corresponding paper (Liu et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', 2019), and they use the mid-training (Phang  et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', 2018) on MNLI for RTE, MRPC, and STS-B while other models (e.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='g.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', BERT, SimVLM, DaVinci, X-FM) do not  use this trick.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' Language Avg.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' is the average score of all the language tasks, while Vision Avg.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' is the average score of six  line evaluation tasks except ImageNet.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' Vision-Language Avg.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' is the average score of all vision-language tasks.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' † are our  reproduced results with the officially released models.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' Uni-Per.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' stands for Uni-Perceiver-MoE (Zhu et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', 2022).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='  4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' Experiments  4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' Pre-training Datasets  We conduct our experiments on several widely used pub-  lic datasets, which consist of two in-domain datasets,  COCO (Lin et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', 2014) and Visual Genome (VG) (Kr-  ishna et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', 2017), and two out-of-domain datasets, SBU  Captions (Ordonez et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', 2011) and Conceptual Captions  (CC) (Sharma et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', 2018).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' Following X-VLM (Zeng et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=',  2021;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' 2022), we also include annotations of objects and  regions from RefCOCO (Yu et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', 2016), Objects365 (Shao  et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', 2019) and OpenImages (Kuznetsova et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', 2018).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='  Since we assume also using uni-modal data, we include  RoBERTa corpus (Liu et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', 2019), C4 datasets (Raffel  et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', 2020) and Imagenet21K (Ridnik et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', 2021).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' All  pre-training datasets are listed in Table 3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='  4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' Implementation Details  Pre-training  Our model is of base size and large size, and  the parameters are listed in Table 5.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' The vision encoder is  initialized with BEiTv2 (Peng et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', 2022).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' The language  encoder is initialized with RoBERTa (Liu et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', 2019).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' The  fusion encoder is trained from scratch.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' X-FM is pre-trained  at image resolution of 224 × 224 with patch size of 16 × 16.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='  Dataset  # Images  # Texts  # Objects  # Regions  COCO  0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='11M  0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='55M  0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='45M VG  0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='10M 2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='0M  3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='7M  SBU  0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='86M  0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='86M CC-3M  2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='9M  2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='9M Objects365  0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='58M 2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='0M OpenImages  1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='7M 4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='2M C4 800GB RoBERTa Corpus 160GB ImageNet-21k  14M Table 3: Statistics of the pre-training datasets.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='  We pre-train X-FMbase for 200K steps with a batch size of  3072 image-text pairs, 3072 images, and 8192 sentences on  32 A100 and pre-train X-FMlarge with the same batch for  160K steps on 64 A100, which takes about six days.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' The  learning rate for both models is warmed-up to 1e−4 in the  first 2500 steps and decayed following a linear schedule.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' We  set the maximum number of text tokens to 30 for image-text  pairs, while that of pure text corpus is set to 128.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' We apply  mixed precision for pre-training.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='  Fine-tuning  We choose widely used downstream tasks  whose details are shown in Appendix B.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' We report overall  Toward Building General Foundation Models for Language, Vision, and Vision-Language Understanding Tasks  Model  # Params  MSCOCO (5K test set)  Flickr30K (1K test set)  MSCOCO (5K test set)  Flickr30K (1K test set)  TR-Fine-Tune  IR-Fine-Tune  TR-Fine-Tune  IR-Fine-Tune  TR-Zero-Shot  IR-Zero-Shot  TR-Zero-Shot  IR-Zero-Shot  R@1/R@5/R@10  R@1/R@5/R@10  R@1/R@5/R@10  R@1/R@5/R@10  R@1/R@5/R@10  R@1/R@5/R@10  R@1/R@5/R@10  R@1/R@5/R@10  ALBEF  210M  73.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='1/91.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='4/96.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='0  56.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='8/81.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='5/89.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='2  94.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='3/99.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='4/99.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='8  82.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='8/96.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='7/98.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='4  –  –  90.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='5/98.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='8/99.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='7  76.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='8/93.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='7/96.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='7  VLMobase  175M  74.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='8/93.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='1/96.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='9  57.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='2/82.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='6/89.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='8  92.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='3/99.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='4/99.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='9  79.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='3/95.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='7/97.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='8  –  –  –  –  VL-BEiT  175M  79.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='5/–/–  61.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='5/–/–  95.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='8/–/–  83.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='9/–/–  –  –  –  –  OmniVL  288M  76.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='8/93.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='6/97.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='3  58.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='5/82.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='6/89.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='5  94.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='9/9.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='6/99.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='9  83.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='4/97.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='0/98.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='6  –  –  –  –  X-VLM  216M  80.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='4/95.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='5/98.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='2  63.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='1/85.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='7/91.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='6  96.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='8/99.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='8/100  86.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='1/97.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='4/98.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='7  70.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='8/92.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='1/96.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='5  55.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='6/82.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='7/90.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='0  85.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='3/97.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='8/99.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='6  71.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='9/93.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='3/96.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='4  X2-VLMbase  255M  80.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='5/95.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='5/97.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='8  62.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='7/84.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='7/90.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='7  97.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='4/99.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='9/100  90.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='0/98.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='6/99.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='3  68.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='4†/92.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='5†/96.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='8†  55.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='2†/82.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='2†/89.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='3†  85.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='1†/99.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='2†/100.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='0†  77.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='3†/95.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='3†/97.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='6†  X-FMbase  284M  81.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='8/96.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='0/98.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='3  64.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='7/86.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='1/91.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='6  97.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='4/100/100  88.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='6/97.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='9/98.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='9  73.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='8/93.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='9/97.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='2  59.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='4/83.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='6/90.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='0  90.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='1/99.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='2/99.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='9  79.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='1/95.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='2/97.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='3  VLMolarge  562M  78.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='2/94.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='4/97.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='4  60.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='6/84.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='4/91.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='0  95.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='3/99.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='9/100  84.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='5/97.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='3/98.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='6  –  –  –  –  X2-VLMlarge  593M  82.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='3/96.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='2/98.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='3  65.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='2/86.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='4/91.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='9  99.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='1/100/100  91.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='1/98.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='6/99.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='4  69.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='7†/93.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='0†/97.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='2†  58.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='3†/83.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='8†/90.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='5†  86.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='8†/98.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='9†/99.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='9†  80.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='5†/96.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='4†/98.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='3†  X-FMlarge  807M  82.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='1/96.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='2/98.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='2  65.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='4/86.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='6/91.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='9  97.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='9/100/100  89.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='4/98.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='2/99.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='1  74.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='4/94.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='1/97.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='3  59.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='4/84.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='4/90.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='7  89.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='7/99.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='1/100  79.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='1/95.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='4/97.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='9  Super-Large Models or Super-Large Datasets  CLIP  490M  –  –  88.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='7/98.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='0/99.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='2  76.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='7/93.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='6/96.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='4  58.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='4/81.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='5/88.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='1  37.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='8/62.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='4/72.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='2  88.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='0/98.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='7/99.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='4  68.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='7/90.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='6/95.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='2  ALIGN  490M  77.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='0/93.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='5/96.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='9  59.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='9/83.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='3/89.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='8  95.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='3/99.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='8/100  84.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='9/97.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='4/98.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='6  58.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='6/83.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='0/89.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='7  45.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='6/69.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='8/78.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='6  88.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='6/98.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='7/99.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='7  75.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='7/93.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='8/96.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='8  Florence  893M  81.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='8/95.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='2/–  63.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='2/85.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='7/–  97.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='2/99.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='9/–  87.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='9/98.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='1/–  64.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='7/85.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='9/–  47.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='2/71.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='4/–  90.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='9/99.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='1/–  76.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='7/93.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='6/–  CoCa  2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='1B  –  –  –  –  66.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='3/86.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='2/91.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='8  51.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='2/74.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='2/82.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='0  92.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='5/99.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='5/99.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='9  80.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='4/95.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='7/97.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='7  BEiT-3  1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='9B  84.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='8/96.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='5/98.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='3  67.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='2/87.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='7/92.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='8  98.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='0/100/100  90.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='3/98.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='7/99.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='5  –  –  94.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='9/99.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='9/100.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='0  81.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='5/95.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='6/97.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='8  X2-VLMlarge  593M  84.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='4/96.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='5/98.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='5  67.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='7/87.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='5/92.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='5  98.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='8/100/100  91.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='8/98.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='6/99.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='5  –  –  –  –  Table 4: Results of text-retrieval (TR) and image-retrieval (IR) on COCO and Flickr30K.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' † denotes our reproduced results  with the officially released models.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' Giant models with over 1B parameters (e.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='g.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', BEiT-3) and models are pre-trained with  over 400M data (e.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='g.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', CLIP and X2-VLMlarge) are in grey since they are not directly comparable with other models.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='  Model  Param  Hidden  Layers  Vision  Text  Fusion  X-FMbase  284  768  12  12  12  X-FMlarge  807  1024  24  24  12  Table 5: Size variants of X-FM.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' All modules consist of  transformer layers.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' Param indicates the parameter number  of transformer layers.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='  performance on eight language tasks from GLUE (Wang  et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', 2019), eight vision tasks following OmniVL (Wang  et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', 2022a), four multi-modal tasks, which are text-  image retrieval on MSCOCO and Flickr, visual question  answering (VQA (Goyal et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', 2017)) and visual reason-  ing (NLVR2 (Suhr et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', 2019b)).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' For image-text retrieval  task, we report both zero-shot results and fine-tuned results.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='  For the ImageNet classification task, we report both linear  evaluation results and fine-tuning results.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' The other vision  tasks are evaluated in the linear evaluation setting.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' All the  other tasks are evaluated in the fine-tuning setting.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' Because  the image resolution differs between pre-training and fine-  tuning, the position parameters are adapted using linear  interpolation.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' For all downstream tasks, we apply random  resize crops and horizontal flips augmentation for the im-  ages during training.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' More details of network architectures  and hyper-parameters setups are given in Appendix C.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='  4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' Comparison with SOTA Foundation Models  We extensively compare the performance of X-FM with  state-of-the-art foundation models on vision, language, and  multi-modal tasks.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' We first compare our model with general  foundation models, including UNIMO-v2 (Li et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', 2021c),  FLAVA (Singh et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', 2021), SimVLM (Wang et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', 2021c),  OFA (Wang et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', 2022b), DaVinci (Diao et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', 2022), Om-  niVL (Wang et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', 2022a), and Uni-Perceiver-MoE (Zhu  et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', 2022).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' We also include comparisons with SOTA  foundation models specifically designed for language, vi-  sion, or vision-language tasks, RoBERTa (Liu et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', 2019),  BEiTv2 (Peng et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', 2022), and X2-VLM (Zeng et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', 2022).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='  There are several observations in Table 2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' First, X-FMbase  (column 11) outperforms all the previous general foundation  models (column 4-10) across almost all tasks by a large mar-  gin, becoming a new and stronger general foundation model.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='  Compared to the previous general foundation models, X-  FMbase improves at least 3.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='2% and the most even to 9.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='7%  on the average of all the reported numbers.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' Second, we com-  pare X-FM with state-of-the-art foundation models specif-  ically designed for language, vision, and vision-language  tasks, RoBERTa, BEiTv2 and X2-VLM.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' We observe that  X-FM is also better than or comparable with the foundation  models at both base and large scale (column 1,2,3 vs 11 and  12,13,14 vs 18).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='  4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' Comparison with SOTA Vision-Language Models  In addition to general foundation models, we also compare  X-FM with state-of-the-art vision-language models.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' The re-  sults are shown in Table 4 and Table 7.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' X-FM demonstrates  its superiority on MSCOCO retrieval and NLVR2, while  achieving competitive performance on Flickr retrieval and  VQA.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' Note that X-FMbase outperforms CLIP, ALIGN and  Florence on image-text retrieval tasks with fewer parame-  ters and much less training data.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' Compared to the recently  released SOTA vision-language model, X2-VLM, X-FM is  much better on image-text retrieval tasks at the zero-shot  setting.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='  Toward Building General Foundation Models for Language, Vision, and Vision-Language Understanding Tasks  X-FMbase  RoBERTa†  S-MLM  S-ITM  wostop  BEiTv2†  woMIM  wBEiTv2 Tokenizer  X2-VLM†  Multi-task  ALL  Task  Eval.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='  1  2  3  4  5  6  7  8  9  10  MNLI  FT  87.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='7  87.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='4  87.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='3  87.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='7  –  –  –  –  87.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='4  87.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='6  CoLA  FT  63.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='2  61.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='6  63.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='6  64.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='2  –  –  –  –  62.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='2  65.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='2  MRPC  FT  90.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='7  92.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='2  91.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='1  90.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='7  –  –  –  –  92.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='0  92.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='5  QQP  FT  91.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='5  91.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='6  91.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='6  91.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='6  –  –  –  –  91.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='6  91.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='6  SST-2  FT  95.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='0  95.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='1  94.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='2  94.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='6  –  –  –  –  94.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='4  95.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='3  QNLI  FT  93.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='1  93.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='0  93.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='2  92.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='5  –  –  –  –  92.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='8  92.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='9  RTE  FT  80.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='9  79.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='1  81.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='6  81.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='2  –  –  –  –  79.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='8  81.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='9  STS-B  FT  90.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='9  90.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='7  90.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='7  90.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='4  –  –  –  –  90.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='1  90.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='8  Language Avg.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='  86.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='6  86.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='4  86.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='7  86.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='6  –  –  –  –  86.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='3  87.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='2  ImageNet  FT  –  –  –  –  85.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='5  84.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='8  85.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='0  –  85.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='0  85.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='3  ImageNet  LE  –  –  –  –  80.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='5  79.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='1  79.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='4  –  79.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='3  81.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='1  Food101  LE  –  –  –  –  88.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='2  86.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='9  87.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='2  –  86.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='9  88.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='7  CIFAR10  LE  –  –  –  –  95.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='3  96.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='6  96.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='5  –  96.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='6  97.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='5  CIFAR100  LE  –  –  –  –  81.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='5  83.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='3  83.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='9  –  84.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='1  86.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='9  Pets  LE  –  –  –  –  93.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='1  88.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='1  88.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='5  –  88.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='2  90.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='7  DTD  LE  –  –  –  –  78.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='4  77.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='7  76.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='9  –  78.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='0  78.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='7  Flowers102  LE  –  –  –  –  95.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='7  94.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='1  94.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='5  –  94.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='2  97.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='1  Vision Avg.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='  –  –  –  –  87.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='3  86.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='3  86.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='5  –  86.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='5  88.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='2  VQAv2  FT  –  78.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='8  78.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='5  78.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='7  –  78.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='3  78.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='2  78.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='0  78.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='2  78.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='6  NLVR2  FT  –  86.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='3  86.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='0  86.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='4  –  85.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='9  85.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='5  86.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='2  86.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='1  86.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='7  Flickr30K TR R@1  ZS  –  88.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='3  87.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='2  87.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='1  –  87.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='1  87.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='2  87.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='7  85.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='0  89.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='3  Flickr30K IR R@1  ZS  –  76.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='6  74.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='9  75.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='8  –  76.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='1  75.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='3  75.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='1  75.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='6  77.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='4  Flickr30K TR R@1  FT  –  97.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='5  97.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='0  97.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='2  –  96.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='4  96.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='7  97.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='0  97.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='0  97.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='7  Flickr30K IR R@1  FT  –  87.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='4  86.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='9  87.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='3  –  86.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='2  86.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='6  86.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='2  86.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='4  87.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='4  COCO TR R@1  ZS  –  72.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='0  72.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='1  70.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='5  –  73.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='0  72.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='1  73.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='2  69.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='9  72.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='8  COCO IR R@1  ZS  –  58.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='4  57.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='1  57.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='7  –  58.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='2  57.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='7  57.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='7  56.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='5  59.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='0  COCO TR R@1  FT  –  81.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='2  80.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='2  80.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='9  –  80.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='6  80.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='1  80.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='3  80.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='0  81.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='2  COCO IR R@1  FT  –  64.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='2  63.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='4  63.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='6  –  63.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='7  63.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='0  63.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='1  63.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='0  64.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='0  Vision-Language Avg.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='  –  79.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='1  78.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='3  78.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='5  –  78.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='6  78.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='2  78.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='5  77.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='8  79.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='4  Table 6: Ablation studies on vision, language, and vision-language tasks.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' We use the same settings as Table 2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' “ALL” for  X-FMbase is trained with the same data under the same settings for pre-training and fine-tuning compared to all the variants.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='  Language Avg.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' is the average of all language tasks, while Vision Avg.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' is the average of all vision tasks.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' Vision-Language  Avg.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' is the average of all vision-language tasks.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' Note that performance of “ALL” is slightly different from X-FMbase in  Table 2, because we use less training steps (160k) for ablation to save the computational resources.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='  4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='5.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' Ablation Study  To verify the contributions of different modules in our frame-  work, we ablate them and evaluate the performance of X-  FM on all downstream tasks.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' The results are shown in  Table 6.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' We first explain several abbreviations in the table.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='  S-MLM means that we only separate the language represen-  tations in the learning IMLM task, while S-ITM means that  language representations for computing ITM and BBP are  separated.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' wostop indicates without stopping the gradients  of all language representations.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' woMIM means that we do  not learn by MIM, while wBEiTv2 tokenizer means that  we learn by MIM with the image tokenizer used in BEiTv2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='  Multi-task is a variation that uses straightforward multi-task  learning to optimize the three encoders in X-FM.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' To make  a fair comparison, we also train RoBERTa, BEiTv2 and  X2-VLM with the same data noted as RoBERTa†, BEiTv2†  and X2-VLM†.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' Note that we also increase the fusion layers  in X2-VLM† to make the parameter sizes comparable to  our models.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' RoBERTa†, BEiTv2† and X2-VLM† all have  slightly better results on average than the official ones.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' From  the results, we have the following observations.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='  First, both designs (stop gradient and masked image mod-  eling) bring improvements, and the combination can make  further improvements on all three downstream tasks (col-  umn 10 vs.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' others).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' Second, without separated language  representations, models always perform worse on language  understanding tasks (column 10 vs.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' 2,3,4).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' Besides, the sep-  arate language representations in the IMLM task on image-  text data are helpful for multi-modal tasks (column 2 vs.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' 4).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='  As we point out in section 1, the fusion encoder can learn  better cross-modal feature alignments by IMLM task from  image-text pairs instead of utilizing text tokens.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' Although S-  ITM shows slight side effects (column 4 vs.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' 3), stopping the  gradients of language representation in the fusion encoder  is necessary to simultaneously achieve strong language un-  derstanding and vision-language understanding capability.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='  Third, the MIM task is useful for vision-language and vi-  sion learning (column 10 vs.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' 6).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' Meanwhile, the targets  in our MIM task are better than the BEiTv2 tokenizer (col-  umn 10 vs.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' 7).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' Four, X-FM is much better than a naive  Toward Building General Foundation Models for Language, Vision, and Vision-Language Understanding Tasks  Method  # Params  VQA  NLVR2  test-dev  test-std  dev  test-P  ALBEF  210M  74.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='5  74.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='7  80.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='2  80.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='5  VLMobase  175M  76.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='6  76.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='9  82.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='8  83.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='3  METER  341M  77.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='7  77.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='6  82.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='3  83.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='1  VL-BEiT  175M  77.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='5  77.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='8  81.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='9  82.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='7  BLIPbase  240M  78.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='2  78.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='2  82.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='5  83.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='1  X-VLM  216M  78.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='1  78.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='1  84.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='2  84.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='2  OFAbase  182M  78.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='0  78.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='1 OmniVL  288M  78.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='3  78.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='4 X2-VLMbase  255M  79.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='2  79.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='3  85.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='9  86.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='1  X-FMbase  284M  79.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='1  79.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='2  86.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='3  86.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='5  VLMolarge  562M  79.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='9  80.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='0  85.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='6  86.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='9  OFAlarge  472M  80.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='3  80.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='5 X2-VLMlarge  593M  80.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='5  80.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='5  87.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='2  87.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='6  X-FMlarge  807M  79.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='5  79.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='6  86.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='2  87.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='8  Super-Large Models or Super-Large Datasets  SimVLMbase  273M  77.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='9  78.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='1  81.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='7  X2-VLMbase  255M  80.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='4  80.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='2  86.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='2  87.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='0  SimVLMlarge  783M  79.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='3  79.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='6  84.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='1  84.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='8  X2-VLMlarge  593M  81.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='9  81.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='8  88.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='7  89.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='4  Florence  893M  80.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='2  80.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='3  –  –  CoCa  2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='1B  82.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='3  82.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='3  86.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='1  87.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='0  BEiT-3  1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='9B  84.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='2  84.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='0  91.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='5  92.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='6  Table 7: Results on VQA and visual reasoning.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' Giant mod-  els with over 1B parameters (e.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='g.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', CoCa and BEiT-3) or  models are pre-trained with over 400M data (e.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='g.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', SimVLM  and X2-VLMlarge) are in grey because they are not directly  comparable with other models.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='  multi-task learning strategy for a foundation model (column  10 vs.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' 8).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' Compared with the straightforward multi-task  strategy, X-FMbase improves an average of 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='9%, 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='7%  and 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='6% on language, vision, and vision-language tasks,  respectively.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' Five, X-FM is also slightly better than foun-  dation models specifically designed for language, vision,  and vision-language tasks with the same training corpus  (column 10 vs.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' 1,5,8).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='  5.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' Conclusion and Limitation  5.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' Conclusion  In this work, we address the problem of how to build a  general foundation model that can perform the best for all  the understanding tasks of language, vision, and vision-  language.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' We propose a general foundation model with two  new and effective training techniques, X-FM, to learn rich  language, vision and vision-language representations at the  same time.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' Experimental results explicitly imply that X-FM  outperforms other general foundation models by a large  margin.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' Moreover, X-FM can even be better or comparable  compared with the SOTA foundation models specifically  designed for language, vision, or vision-language under-  standing tasks.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='  5.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' Limitation  Like most existing work on foundation models, the entire  project consumed over 5 A100 GPU years on a computing  cluster with high electricity costs, although we only tested  base and large models.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' There is still potential for efficiency  improvement through sparse attention (Zaheer et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', 2020)  or the lottery ticket hypothesis (Frankle & Carbin, 2018).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='  We will explore the techniques to improve the training effi-  ciency and reduce the carbon footprint so that we can adhere  to the proposals on “green” deep learning (Schwartz et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=',  2020;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' Xu et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', 2021).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='  Due to considerations of fair comparisons and computa-  tional resources, we did not try super-large models which  use at least 1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='9B or more parameters like BEITv3 (Wang  et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', 2022d), CoCa (Yu et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', 2022) and PaLI (Chen et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=',  2022).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' However, scalability is also an important factor for  foundation models.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' We leave the investigations to future  work.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='  References  Agirre, E.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', Màrquez, L.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', and Wicentowski, R.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' (eds.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=').' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' Pro-  ceedings of the Fourth International Workshop on Seman-  tic Evaluations (SemEval-2007), Prague, Czech Republic,  2007.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' Association for Computational Linguistics.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' URL  https://aclanthology.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='org/S07-1000.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='  Alayrac, J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='-B.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', Donahue, J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', Luc, P.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', Miech, A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', Barr, I.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=',  Hasson, Y.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', Lenc, K.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', Mensch, A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', Millican, K.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', Reynolds,  M.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' Flamingo: a visual language model for few-shot  learning.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' arXiv preprint arXiv:2204.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='14198, 2022.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='  Bao, H.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', Dong, L.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', and Wei, F.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' BEiT: Bert pre-training of  image transformers.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' arXiv preprint, 2021.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='  Bao, H.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', Wang, W.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', Dong, L.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', and Wei, F.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='  Vl-beit:  Generative vision-language pretraining.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' arXiv preprint  arXiv:2206.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='01127, 2022.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='  Bentivogli, L.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', Clark, P.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', Dagan, I.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', and Giampiccolo, D.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='  The fifth pascal recognizing textual entailment challenge.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='  In TAC, 2009.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='  Bingel, J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' and Søgaard, A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' Identifying beneficial task re-  lations for multi-task learning in deep neural networks.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='  arXiv preprint arXiv:1702.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='08303, 2017.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='  Bossard, L.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', Guillaumin, M.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', and Gool, L.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' V.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' Food-101–  mining discriminative components with random forests.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='  In European conference on computer vision, pp.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' 446–461.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='  Springer, 2014.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='  Brown, T.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' B.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', Mann, B.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', Ryder, N.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', Subbiah, M.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', Kaplan,  J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', Dhariwal, P.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', Neelakantan, A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', Shyam, P.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', Sastry, G.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=',  Toward Building General Foundation Models for Language, Vision, and Vision-Language Understanding Tasks  Askell, A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', Agarwal, S.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', Herbert-Voss, A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', Krueger, G.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=',  Henighan, T.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', Child, R.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', Ramesh, A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', Ziegler, D.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' M.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=',  Wu, J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', Winter, C.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', Hesse, C.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', Chen, M.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', Sigler, E.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=',  Litwin, M.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', Gray, S.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', Chess, B.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', Clark, J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', Berner,  C.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', McCandlish, S.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', Radford, A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', Sutskever, I.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', and  Amodei, D.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='  Language models are few-shot learners.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='  In Larochelle, H.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', Ranzato, M.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', Hadsell, R.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', Balcan,  M.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', and Lin, H.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' (eds.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' ), Advances in Neural Information  Processing Systems 33: Annual Conference on Neural  Information Processing Systems 2020, NeurIPS 2020,  December 6-12, 2020, virtual, 2020.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='  URL https:  //proceedings.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='neurips.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='cc/paper/2020/hash/  1457c0d6bfcb4967418bfb8ac142f64a-Abstract.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='  html.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='  Caron, M.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', Touvron, H.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', Misra, I.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', Jégou, H.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', Mairal, J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=',  Bojanowski, P.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', and Joulin, A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' Emerging properties in  self-supervised vision transformers.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' In Proceedings of  the IEEE/CVF International Conference on Computer  Vision, pp.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' 9650–9660, 2021.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='  Chen, X.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', Wang, X.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', Changpinyo, S.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', Piergiovanni, A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=',  Padlewski, P.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', Salz, D.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', Goodman, S.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', Grycner, A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=',  Mustafa, B.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', Beyer, L.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='  Pali: A jointly-scaled  multilingual language-image model.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='  arXiv preprint  arXiv:2209.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='06794, 2022.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='  Chen, Y.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='-C.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', Li, L.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', Yu, L.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', El Kholy, A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', Ahmed, F.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', Gan,  Z.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', Cheng, Y.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', and Liu, J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' UNITER: Universal image-  text representation learning.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' In European Conference on  Computer Vision (ECCV), 2020.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='  Cho, J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', Lei, J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', Tan, H.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', and Bansal, M.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='  Unifying  vision-and-language tasks via text generation.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' In Meila,  M.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' and Zhang, T.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' (eds.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' ), Proceedings of the 38th In-  ternational Conference on Machine Learning, ICML  2021, 18-24 July 2021, Virtual Event, volume 139 of  Proceedings of Machine Learning Research, pp.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' 1931–  1942.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' PMLR, 2021.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' URL http://proceedings.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='mlr.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='  press/v139/cho21a.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='html.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='  Cimpoi, M.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', Maji, S.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', Kokkinos, I.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', Mohamed, S.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', and  Vedaldi, A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' Describing textures in the wild.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' In 2014  IEEE Conference on Computer Vision and Pattern Recog-  nition, CVPR 2014, Columbus, OH, USA, June 23-28,  2014, pp.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' 3606–3613.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' IEEE Computer Society, 2014.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' doi:  10.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='1109/CVPR.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='2014.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='461.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' URL https://doi.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='org/10.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='  1109/CVPR.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='2014.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='461.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='  Clark, K.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', Luong, M.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', Le, Q.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' V.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', and Manning, C.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' D.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='  ELECTRA: pre-training text encoders as discriminators  rather than generators.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='  In 8th International Confer-  ence on Learning Representations, ICLR 2020, Addis  Ababa, Ethiopia, April 26-30, 2020.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' OpenReview.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='net,  2020.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' URL https://openreview.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='net/forum?' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='id=  r1xMH1BtvB.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='  Dagan, I.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', Glickman, O.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', and Magnini, B.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' The pascal recog-  nising textual entailment challenge.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' In Machine Learning  Challenges Workshop, pp.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' 177–190.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' Springer, 2005.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='  Devlin, J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', Chang, M.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='-W.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', Lee, K.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', and Toutanova, K.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' BERT:  Pre-training of deep bidirectional transformers for lan-  guage understanding.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' In Proceedings of the 2019 Confer-  ence of the North American Chapter of the Association for  Computational Linguistics: Human Language Technolo-  gies, Volume 1 (Long and Short Papers), pp.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' 4171–4186,  Minneapolis, Minnesota, 2019.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' Association for Compu-  tational Linguistics.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' doi: 10.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='18653/v1/N19-1423.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' URL  https://aclanthology.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='org/N19-1423.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='  Diao, S.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', Zhou, W.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', Zhang, X.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', and Wang, J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' Prefix lan-  guage models are unified modal learners.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' arXiv preprint  arXiv:2206.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='07699, 2022.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='  Dolan, W.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' B.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' and Brockett, C.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' Automatically construct-  ing a corpus of sentential paraphrases.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='  In Proceed-  ings of the Third International Workshop on Paraphras-  ing (IWP2005), 2005.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' URL https://aclanthology.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='  org/I05-5002.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='  Dosovitskiy, A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', Beyer, L.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', Kolesnikov, A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', Weissenborn,  D.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', Zhai, X.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', Unterthiner, T.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', Dehghani, M.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', Minderer,  M.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', Heigold, G.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', Gelly, S.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', Uszkoreit, J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', and Houlsby, N.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='  An image is worth 16x16 words: Transformers for image  recognition at scale.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' In 9th International Conference on  Learning Representations, ICLR 2021, Virtual Event, Aus-  tria, May 3-7, 2021.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' OpenReview.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='net, 2021.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' URL https:  //openreview.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='net/forum?' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='id=YicbFdNTTy.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='  Frankle, J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' and Carbin, M.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' The lottery ticket hypothesis:  Finding sparse, trainable neural networks.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' arXiv preprint  arXiv:1803.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='03635, 2018.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='  Giampiccolo, D.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', Magnini, B.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', Dagan, I.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', and Dolan, B.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' The  third PASCAL recognizing textual entailment challenge.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='  In Proceedings of the ACL-PASCAL Workshop on Textual  Entailment and Paraphrasing, pp.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' 1–9, Prague, 2007.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='  Association for Computational Linguistics.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' URL https:  //aclanthology.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='org/W07-1401.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='  Goyal, Y.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', Khot, T.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', Summers-Stay, D.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', Batra, D.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', and  Parikh, D.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' Making the V in VQA matter: Elevating  the role of image understanding in visual question an-  swering.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' In 2017 IEEE Conference on Computer Vi-  sion and Pattern Recognition, CVPR 2017, Honolulu,  HI, USA, July 21-26, 2017, pp.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' 6325–6334.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' IEEE Com-  puter Society, 2017.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' doi: 10.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='1109/CVPR.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='2017.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='670.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' URL  https://doi.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='org/10.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='1109/CVPR.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='2017.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='670.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='  Haim, R.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' B.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', Dagan, I.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', Dolan, B.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', Ferro, L.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', Giampiccolo,  D.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', Magnini, B.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', and Szpektor, I.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' The second pascal recog-  nising textual entailment challenge.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' In Proceedings of the  Second PASCAL Challenges Workshop on Recognising  Textual Entailment, volume 7, 2006.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='  Toward Building General Foundation Models for Language, Vision, and Vision-Language Understanding Tasks  He, K.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', Fan, H.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', Wu, Y.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', Xie, S.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', and Girshick, R.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' B.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' Mo-  mentum contrast for unsupervised visual representation  learning.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' In 2020 IEEE/CVF Conference on Computer  Vision and Pattern Recognition, CVPR 2020, Seattle, WA,  USA, June 13-19, 2020, pp.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' 9726–9735.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' IEEE, 2020.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='  doi: 10.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='1109/CVPR42600.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='2020.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='00975.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' URL https:  //doi.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='org/10.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='1109/CVPR42600.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='2020.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='00975.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='  He, K.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', Chen, X.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', Xie, S.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', Li, Y.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', Dollár, P.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', and Girshick,  R.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' Masked autoencoders are scalable vision learners.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' In  Proceedings of the IEEE/CVF Conference on Computer  Vision and Pattern Recognition, pp.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' 16000–16009, 2022.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='  He, P.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', Liu, X.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', Gao, J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', and Chen, W.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' Deberta: decoding-  enhanced bert with disentangled attention.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' In 9th Interna-  tional Conference on Learning Representations, ICLR  2021, Virtual Event, Austria, May 3-7, 2021.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' Open-  Review.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='net, 2021.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' URL https://openreview.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='net/  forum?' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='id=XPZIaotutsD.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='  Iyer, S.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', Dandekar, N.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', Csernai, K.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' First quora dataset  release: Question pairs.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' data.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' quora.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' com, 2017.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='  Jia, C.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', Yang, Y.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', Xia, Y.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', Chen, Y.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', Parekh, Z.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', Pham, H.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=',  Le, Q.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' V.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', Sung, Y.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', Li, Z.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', and Duerig, T.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' Scaling up  visual and vision-language representation learning with  noisy text supervision.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' In Meila, M.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' and Zhang, T.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' (eds.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' ),  Proceedings of the 38th International Conference on Ma-  chine Learning, ICML 2021, 18-24 July 2021, Virtual  Event, volume 139 of Proceedings of Machine Learning  Research, pp.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' 4904–4916.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' PMLR, 2021.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' URL http:  //proceedings.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='mlr.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='press/v139/jia21b.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='html.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='  Joshi, M.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', Chen, D.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', Liu, Y.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', Weld, D.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' S.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', Zettlemoyer,  L.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', and Levy, O.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' Spanbert: Improving pre-training by  representing and predicting spans.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' Transactions of the As-  sociation for Computational Linguistics, 8:64–77, 2020.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='  Karpathy, A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' and Li, F.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' Deep visual-semantic alignments  for generating image descriptions.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' In IEEE Conference  on Computer Vision and Pattern Recognition, CVPR  2015, Boston, MA, USA, June 7-12, 2015, pp.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' 3128–3137.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='  IEEE Computer Society, 2015.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' doi: 10.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='1109/CVPR.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='2015.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='  7298932.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='  URL https://doi.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='org/10.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='1109/CVPR.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='  2015.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='7298932.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='  Krishna, R.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', Zhu, Y.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', Groth, O.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', Johnson, J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', Hata, K.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=',  Kravitz, J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', Chen, S.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', Kalantidis, Y.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', Li, L.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='-J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', Shamma,  D.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' Visual genome: Connecting language and  vision using crowdsourced dense image annotations.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' In-  ternational Journal of Computer Vision (IJCV), 2017.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='  Krizhevsky, A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', Hinton, G.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' Learning multiple layers  of features from tiny images.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' 2009.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='  Kuznetsova, A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', Rom, H.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', Alldrin, N.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', Uijlings, J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', Krasin,  I.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', Pont-Tuset, J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', Kamali, S.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', Popov, S.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', Malloci, M.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=',  Kolesnikov, A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' The open images dataset v4: Unified  image classification, object detection, and visual relation-  ship detection at scale.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' arXiv preprint arXiv:1811.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='00982,  2018.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' URL https://arxiv.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='org/abs/1811.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='00982.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='  Lan, Z.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', Chen, M.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', Goodman, S.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', Gimpel, K.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', Sharma, P.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=',  and Soricut, R.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' ALBERT: A lite BERT for self-supervised  learning of language representations.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' In 8th International  Conference on Learning Representations, ICLR 2020, Ad-  dis Ababa, Ethiopia, April 26-30, 2020.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' OpenReview.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='net,  2020.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' URL https://openreview.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='net/forum?' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='id=  H1eA7AEtvS.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='  Lewis, M.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', Liu, Y.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', Goyal, N.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', Ghazvininejad, M.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', Mo-  hamed, A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', Levy, O.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', Stoyanov, V.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', and Zettlemoyer, L.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='  BART: Denoising sequence-to-sequence pre-training for  natural language generation, translation, and comprehen-  sion.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' In Proceedings of the 58th Annual Meeting of the As-  sociation for Computational Linguistics, pp.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' 7871–7880,  Online, 2020.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' Association for Computational Linguis-  tics.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' doi: 10.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='18653/v1/2020.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='acl-main.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='703.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' URL https:  //aclanthology.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='org/2020.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='acl-main.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='703.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='  Li, J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', Selvaraju, R.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' R.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', Gotmare, A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' D.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', Joty, S.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', Xiong,  C.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', and Hoi, S.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' Align before fuse: Vision and language  representation learning with momentum distillation.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' In  Conference on Neural Information Processing Systems  (NeurIPS), 2021a.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='  Li, J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', Li, D.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', Xiong, C.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', and Hoi, S.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='  Blip:  Boot-  strapping language-image pre-training for unified vision-  language understanding and generation.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' arXiv preprint  arXiv:2201.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='12086, 2022.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='  Li, W.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', Gao, C.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', Niu, G.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', Xiao, X.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', Liu, H.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', Liu, J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', Wu,  H.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', and Wang, H.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' UNIMO: Towards unified-modal un-  derstanding and generation via cross-modal contrastive  learning.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' In Proceedings of the 59th Annual Meeting of  the Association for Computational Linguistics and the  11th International Joint Conference on Natural Language  Processing (Volume 1: Long Papers), pp.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' 2592–2607,  Online, 2021b.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' Association for Computational Linguis-  tics.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' doi: 10.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='18653/v1/2021.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='acl-long.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='202.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' URL https:  //aclanthology.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='org/2021.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='acl-long.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='202.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='  Li, W.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', Gao, C.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', Niu, G.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', Xiao, X.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', Liu, H.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', Liu, J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', Wu,  H.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', and Wang, H.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' UNIMO: Towards unified-modal un-  derstanding and generation via cross-modal contrastive  learning.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' In Proceedings of the 59th Annual Meeting of  the Association for Computational Linguistics and the  11th International Joint Conference on Natural Language  Processing (Volume 1: Long Papers), pp.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' 2592–2607,  Online, 2021c.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' Association for Computational Linguis-  tics.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' doi: 10.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='18653/v1/2021.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='acl-long.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='202.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' URL https:  //aclanthology.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='org/2021.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='acl-long.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='202.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='  Toward Building General Foundation Models for Language, Vision, and Vision-Language Understanding Tasks  Li, X.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', Yin, X.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', Li, C.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', Zhang, P.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', Hu, X.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', Zhang, L.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', Wang,  L.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', Hu, H.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', Dong, L.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', Wei, F.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='  Oscar: Object-  semantics aligned pre-training for vision-language tasks.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='  In European Conference on Computer Vision (ECCV),  2020.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='  Lin, T.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='-Y.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', Maire, M.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', Belongie, S.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', Hays, J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', Perona, P.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=',  Ramanan, D.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', Dollár, P.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', and Zitnick, C.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' L.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' Microsoft  COCO: Common objects in context.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' In European Con-  ference on Computer Vision (ECCV), 2014.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='  Liu, Y.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', Ott, M.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', Goyal, N.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', Du, J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', Joshi, M.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', Chen, D.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=',  Levy, O.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', Lewis, M.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', Zettlemoyer, L.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', and Stoyanov,  V.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='  RoBERTa: A robustly optimized bert pretraining  approach.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' arXiv preprint, 2019.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='  Lu, J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', Batra, D.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', Parikh, D.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', and Lee, S.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' Vilbert: Pre-  training task-agnostic visiolinguistic representations  for vision-and-language tasks.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='  In Wallach, H.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' M.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=',  Larochelle, H.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', Beygelzimer, A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', d’Alché-Buc, F.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=',  Fox, E.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' B.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', and Garnett, R.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' (eds.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' ), Advances in Neural  Information Processing Systems 32:  Annual Con-  ference on Neural Information Processing Systems  2019, NeurIPS 2019, December 8-14, 2019, Vancou-  ver, BC, Canada, pp.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' 13–23, 2019.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='  URL https:  //proceedings.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='neurips.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='cc/paper/2019/hash/  c74d97b01eae257e44aa9d5bade97baf-Abstract.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='  html.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='  Nilsback, M.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='-E.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' and Zisserman, A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' Automated flower clas-  sification over a large number of classes.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' In 2008 Sixth  Indian Conference on Computer Vision, Graphics & Im-  age Processing, pp.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' 722–729.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' IEEE, 2008.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='  Ordonez, V.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', Kulkarni, G.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', and Berg, T.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' L.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' Im2text: Describ-  ing images using 1 million captioned photographs.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' In  Shawe-Taylor, J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', Zemel, R.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' S.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', Bartlett, P.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' L.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', Pereira, F.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='  C.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' N.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', and Weinberger, K.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' Q.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' (eds.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' ), Advances in Neural  Information Processing Systems 24: 25th Annual Confer-  ence on Neural Information Processing Systems 2011.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='  Proceedings of a meeting held 12-14 December 2011,  Granada, Spain, pp.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' 1143–1151, 2011.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' URL https:  //proceedings.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='neurips.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='cc/paper/2011/hash/  5dd9db5e033da9c6fb5ba83c7a7ebea9-Abstract.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='  html.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='  Parkhi, O.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' M.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', Vedaldi, A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', Zisserman, A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', and Jawahar,  C.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' V.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='  Cats and dogs.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='  In 2012 IEEE Conference on  Computer Vision and Pattern Recognition, Providence, RI,  USA, June 16-21, 2012, pp.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' 3498–3505.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' IEEE Computer  Society, 2012.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' doi: 10.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='1109/CVPR.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='2012.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='6248092.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' URL  https://doi.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='org/10.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='1109/CVPR.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='2012.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='6248092.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='  Peng, Z.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', Dong, L.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', Bao, H.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', Ye, Q.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', and Wei, F.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' Beit v2:  Masked image modeling with vector-quantized visual  tokenizers.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' arXiv preprint arXiv:2208.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='06366, 2022.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='  Phang, J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', Févry, T.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', and Bowman, S.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' R.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' Sentence encoders  on stilts: Supplementary training on intermediate labeled-  data tasks.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' ArXiv, abs/1811.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='01088, 2018.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='  Radford, A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', Kim, J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' W.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', Hallacy, C.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', Ramesh, A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', Goh, G.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=',  Agarwal, S.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', Sastry, G.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', Askell, A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', Mishkin, P.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', Clark,  J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', Krueger, G.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', and Sutskever, I.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' Learning transferable  visual models from natural language supervision.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='  In  Meila, M.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' and Zhang, T.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' (eds.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' ), Proceedings of the 38th  International Conference on Machine Learning, ICML  2021, 18-24 July 2021, Virtual Event, volume 139 of  Proceedings of Machine Learning Research, pp.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' 8748–  8763.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' PMLR, 2021.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' URL http://proceedings.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='mlr.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='  press/v139/radford21a.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='html.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='  Raffel, C.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', Shazeer, N.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', Roberts, A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', Lee, K.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', Narang, S.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=',  Matena, M.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', Zhou, Y.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', Li, W.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', and Liu, P.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' Exploring the  limits of transfer learning with a unified text-to-text trans-  former.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' Journal of Machine Learning Research (JMLR),  2020.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='  Rajpurkar, P.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', Zhang, J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', Lopyrev, K.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', and Liang, P.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' SQuAD:  100,000+ questions for machine comprehension of text.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='  In Proceedings of the 2016 Conference on Empirical  Methods in Natural Language Processing, pp.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' 2383–  2392, Austin, Texas, 2016.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' Association for Computa-  tional Linguistics.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' doi: 10.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='18653/v1/D16-1264.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' URL  https://aclanthology.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='org/D16-1264.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='  Rezatofighi, H.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', Tsoi, N.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', Gwak, J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', Sadeghian, A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', Reid,  I.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' D.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', and Savarese, S.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' Generalized intersection over  union: A metric and a loss for bounding box regression.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='  In IEEE Conference on Computer Vision and Pattern  Recognition, CVPR 2019, Long Beach, CA, USA, June  16-20, 2019, pp.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' 658–666.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' Computer Vision Foundation /  IEEE, 2019.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' doi: 10.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='1109/CVPR.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='2019.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='00075.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='  Ridnik, T.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', Ben-Baruch, E.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', Noy, A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', and Zelnik-Manor, L.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='  Imagenet-21k pretraining for the masses, 2021.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='  Russakovsky, O.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', Deng, J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', Su, H.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', Krause, J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', Satheesh, S.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=',  Ma, S.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', Huang, Z.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', Karpathy, A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', Khosla, A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', Bernstein,  M.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' Imagenet large scale visual recognition chal-  lenge.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' International journal of computer vision, 115(3):  211–252, 2015.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='  Schwartz, R.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', Dodge, J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', Smith, N.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', and Etzioni, O.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' Green  ai.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' Communications of the ACM, 63(12):54–63, 2020.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='  Shao, S.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', Li, Z.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', Zhang, T.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', Peng, C.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', Yu, G.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', Zhang, X.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', Li,  J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', and Sun, J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' Objects365: A large-scale, high-quality  dataset for object detection.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' In 2019 IEEE/CVF Interna-  tional Conference on Computer Vision, ICCV 2019, Seoul,  Korea (South), October 27 - November 2, 2019, pp.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' 8429–  8438.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' IEEE, 2019.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' doi: 10.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='1109/ICCV.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='2019.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='00852.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' URL  https://doi.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='org/10.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='1109/ICCV.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='2019.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='00852.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='  Toward Building General Foundation Models for Language, Vision, and Vision-Language Understanding Tasks  Sharma, P.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', Ding, N.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', Goodman, S.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', and Soricut, R.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' Con-  ceptual captions: A cleaned, hypernymed, image alt-text  dataset for automatic image captioning.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' In Proceedings  of the 56th Annual Meeting of the Association for Compu-  tational Linguistics (Volume 1: Long Papers), pp.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' 2556–  2565, Melbourne, Australia, 2018.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' Association for Com-  putational Linguistics.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' doi: 10.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='18653/v1/P18-1238.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' URL  https://aclanthology.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='org/P18-1238.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='  Singh, A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', Hu, R.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', Goswami, V.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', Couairon, G.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', Galuba,  W.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', Rohrbach, M.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', and Kiela, D.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' Flava: A foundational  language and vision alignment model.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' ArXiv preprint,  abs/2112.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='04482, 2021.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='  URL https://arxiv.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='org/  abs/2112.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='04482.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='  Socher, R.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', Perelygin, A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', Wu, J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', Chuang, J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', Manning,  C.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' D.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', Ng, A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', and Potts, C.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' Recursive deep models for  semantic compositionality over a sentiment treebank.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' In  Proceedings of the 2013 Conference on Empirical Meth-  ods in Natural Language Processing, pp.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' 1631–1642,  Seattle, Washington, USA, 2013.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' Association for Com-  putational Linguistics.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' URL https://aclanthology.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='  org/D13-1170.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='  Suhr, A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', Zhou, S.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', Zhang, A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', Zhang, I.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', Bai, H.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', and  Artzi, Y.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' A corpus for reasoning about natural language  grounded in photographs.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' In Proceedings of the 57th  Annual Meeting of the Association for Computational  Linguistics, pp.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' 6418–6428, Florence, Italy, 2019a.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' As-  sociation for Computational Linguistics.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' doi: 10.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='18653/  v1/P19-1644.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='  URL https://aclanthology.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='org/  P19-1644.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='  Suhr, A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', Zhou, S.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', Zhang, A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', Zhang, I.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', Bai, H.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', and  Artzi, Y.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' A corpus for reasoning about natural language  grounded in photographs.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' In Proceedings of the 57th  Annual Meeting of the Association for Computational  Linguistics, pp.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' 6418–6428, Florence, Italy, 2019b.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' As-  sociation for Computational Linguistics.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' doi: 10.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='18653/  v1/P19-1644.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='  URL https://aclanthology.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='org/  P19-1644.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='  Sun, Y.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', Wang, S.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', Li, Y.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', Feng, S.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', Chen, X.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', Zhang, H.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=',  Tian, X.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', Zhu, D.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', Tian, H.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', and Wu, H.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' Ernie: Enhanced  representation through knowledge integration.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='  arXiv  preprint arXiv:1904.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='09223, 2019.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='  Tan, H.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' and Bansal, M.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' LXMERT: Learning cross-modality  encoder representations from transformers.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='  In Pro-  ceedings of the 2019 Conference on Empirical Meth-  ods in Natural Language Processing and the 9th In-  ternational Joint Conference on Natural Language Pro-  cessing (EMNLP-IJCNLP), pp.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' 5100–5111, Hong Kong,  China, 2019a.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' Association for Computational Linguis-  tics.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='  doi: 10.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='18653/v1/D19-1514.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='  URL https://  aclanthology.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='org/D19-1514.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='  Tan, H.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' and Bansal, M.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' LXMERT: Learning cross-modality  encoder representations from transformers.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='  In Pro-  ceedings of the 2019 Conference on Empirical Meth-  ods in Natural Language Processing and the 9th In-  ternational Joint Conference on Natural Language Pro-  cessing (EMNLP-IJCNLP), pp.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' 5100–5111, Hong Kong,  China, 2019b.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' Association for Computational Linguis-  tics.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='  doi: 10.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='18653/v1/D19-1514.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='  URL https://  aclanthology.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='org/D19-1514.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='  Touvron, H.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', Cord, M.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', Douze, M.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', Massa, F.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', Sablay-  rolles, A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', and Jégou, H.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='  Training data-efficient im-  age transformers & distillation through attention.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='  In  Meila, M.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' and Zhang, T.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' (eds.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' ), Proceedings of the 38th  International Conference on Machine Learning, ICML  2021, 18-24 July 2021, Virtual Event, volume 139 of  Proceedings of Machine Learning Research, pp.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' 10347–  10357.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' PMLR, 2021.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='  URL http://proceedings.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='  mlr.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='press/v139/touvron21a.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='html.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='  Vaswani, A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', Shazeer, N.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', Parmar, N.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', Uszkoreit, J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=',  Jones, L.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', Gomez, A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' N.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', Kaiser, L.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', and Polosukhin,  I.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='  Attention is all you need.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='  In Guyon, I.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', von  Luxburg, U.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', Bengio, S.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', Wallach, H.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' M.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', Fergus,  R.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', Vishwanathan, S.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' V.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' N.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', and Garnett, R.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' (eds.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' ),  Advances in Neural Information Processing Systems  30: Annual Conference on Neural Information Pro-  cessing Systems 2017, December 4-9, 2017, Long  Beach, CA, USA, pp.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' 5998–6008, 2017.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' URL https:  //proceedings.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='neurips.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='cc/paper/2017/hash/  3f5ee243547dee91fbd053c1c4a845aa-Abstract.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='  html.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='  Wang, A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', Singh, A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', Michael, J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', Hill, F.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', Levy, O.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', and  Bowman, S.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' R.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' GLUE: A multi-task benchmark and  analysis platform for natural language understanding.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' In  7th International Conference on Learning Representa-  tions, ICLR 2019, New Orleans, LA, USA, May 6-9, 2019.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='  OpenReview.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='net, 2019.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' URL https://openreview.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='  net/forum?' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='id=rJ4km2R5t7.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='  Wang, J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', Chen, D.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', Wu, Z.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', Luo, C.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', Zhou, L.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', Zhao, Y.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=',  Xie, Y.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', Liu, C.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', Jiang, Y.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='-G.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', and Yuan, L.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' Omnivl:  One foundation model for image-language and video-  language tasks.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' arXiv preprint arXiv:2209.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='07526, 2022a.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='  Wang, P.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', Yang, A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', Men, R.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', Lin, J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', Bai, S.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', Li, Z.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', Ma, J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=',  Zhou, C.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', Zhou, J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', and Yang, H.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' Ofa: Unifying architec-  tures, tasks, and modalities through a simple sequence-  to-sequence learning framework.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' In International Con-  ference on Machine Learning, pp.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' 23318–23340.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' PMLR,  2022b.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='  Wang, P.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', Yang, A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', Men, R.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', Lin, J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', Bai, S.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', Li, Z.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=',  Ma, J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', Zhou, C.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', Zhou, J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', and Yang, H.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='  Unifying  architectures, tasks, and modalities through a simple  Toward Building General Foundation Models for Language, Vision, and Vision-Language Understanding Tasks  sequence-to-sequence learning framework.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' arXiv preprint  arXiv:2202.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='03052, 2022c.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='  Wang, W.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', Tran, D.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', and Feiszli, M.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' What makes training  multi-modal classification networks hard?' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' In Proceedings  of the IEEE/CVF Conference on Computer Vision and  Pattern Recognition, pp.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' 12695–12705, 2020.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='  Wang, W.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', Bao, H.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', Dong, L.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', and Wei, F.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' Vlmo: Unified  vision-language pre-training with mixture-of-modality-  experts.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' ArXiv preprint, abs/2111.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='02358, 2021a.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' URL  https://arxiv.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='org/abs/2111.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='02358.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='  Wang, W.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', Bao, H.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', Dong, L.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', Bjorck, J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', Peng, Z.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', Liu,  Q.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', Aggarwal, K.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', Mohammed, O.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' K.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', Singhal, S.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', Som,  S.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' Image as a foreign language: Beit pretraining  for all vision and vision-language tasks.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' arXiv preprint  arXiv:2208.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='10442, 2022d.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='  Wang, Z.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', Yu, J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', Yu, A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' W.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', Dai, Z.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', Tsvetkov, Y.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', and Cao,  Y.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' Simvlm: Simple visual language model pretraining  with weak supervision.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' CoRR, abs/2108.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='10904, 2021b.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='  Wang, Z.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', Yu, J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', Yu, A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' W.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', Dai, Z.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', Tsvetkov, Y.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', and Cao,  Y.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' Simvlm: Simple visual language model pretraining  with weak supervision.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' arXiv preprint, 2021c.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='  Warstadt, A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', Singh, A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', and Bowman, S.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' R.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' Neural network  acceptability judgments.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' Transactions of the Association  for Computational Linguistics, 7:625–641, 2019.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' doi:  10.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='1162/tacl_a_00290.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' URL https://aclanthology.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='  org/Q19-1040.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='  Wei, C.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', Fan, H.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', Xie, S.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', Wu, C.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='-Y.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', Yuille, A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', and Feicht-  enhofer, C.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' Masked feature prediction for self-supervised  visual pre-training.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' In Proceedings of the IEEE/CVF Con-  ference on Computer Vision and Pattern Recognition, pp.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='  14668–14678, 2022a.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='  Wei, L.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', Xie, L.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', Zhou, W.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', Li, H.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', and Tian, Q.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' Mvp:  Multimodality-guided visual pre-training.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' arXiv preprint  arXiv:2203.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='05175, 2022b.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='  Williams, A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', Nangia, N.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', and Bowman, S.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='  A broad-  coverage challenge corpus for sentence understanding  through inference.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='  In Proceedings of the 2018 Con-  ference of the North American Chapter of the Associ-  ation for Computational Linguistics: Human Language  Technologies, Volume 1 (Long Papers), pp.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' 1112–1122,  New Orleans, Louisiana, 2018.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' Association for Compu-  tational Linguistics.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' doi: 10.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='18653/v1/N18-1101.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' URL  https://aclanthology.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='org/N18-1101.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='  Xu, J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', Zhou, W.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', Fu, Z.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', Zhou, H.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', and Li, L.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' A survey  on green deep learning.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' ArXiv preprint, abs/2111.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='05193,  2021.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' URL https://arxiv.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='org/abs/2111.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='05193.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='  Yu, J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', Wang, Z.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', Vasudevan, V.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', Yeung, L.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', Seyedhosseini,  M.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', and Wu, Y.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' Coca: Contrastive captioners are image-  text foundation models.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' arXiv preprint arXiv:2205.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='01917,  2022.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='  Yu, L.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', Poirson, P.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', Yang, S.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', Berg, A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' C.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', and Berg, T.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' L.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='  Modeling context in referring expressions.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' In European  Conference on Computer Vision, pp.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' 69–85.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' Springer,  2016.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='  Yuan, L.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', Chen, D.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', Chen, Y.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='-L.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', Codella, N.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', Dai, X.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', Gao,  J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', Hu, H.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', Huang, X.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', Li, B.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', Li, C.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', Liu, C.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', Liu, M.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=',  Liu, Z.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', Lu, Y.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', Shi, Y.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', Wang, L.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', Wang, J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', Xiao, B.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=',  Xiao, Z.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', Yang, J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', Zeng, M.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', Zhou, L.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', and Zhang, P.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='  Florence: A new foundation model for computer vision.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='  arXiv preprint, 2021.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='  Zaheer, M.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', Guruganesh, G.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', Dubey, K.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', Ainslie, J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', Al-  berti, C.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', Ontanon, S.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', Pham, P.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', Ravula, A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', Wang, Q.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=',  Yang, L.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' Big bird: Transformers for longer se-  quences.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' Advances in Neural Information Processing  Systems, 33:17283–17297, 2020.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='  Zeng, Y.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', Zhang, X.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', and Li, H.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='  Multi-grained vision  language pre-training: Aligning texts with visual con-  cepts.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='  ArXiv preprint, abs/2111.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='08276, 2021.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='  URL  https://arxiv.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='org/abs/2111.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='08276.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='  Zeng, Y.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', Zhang, X.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', Li, H.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', Wang, J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', Zhang, J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', and Zhou,  W.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' X2-vlm: All-in-one pre-trained model for vision-  language tasks.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' arXiv preprint arXiv:2211.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='12402, 2022.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='  Zhang, P.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', Li, X.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', Hu, X.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', Yang, J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', Zhang, L.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', Wang, L.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=',  Choi, Y.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', and Gao, J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' VinVL: Revisiting visual repre-  sentations in vision-language models.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' In Conference on  Computer Vision and Pattern Recognition (CVPR), 2021.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='  Zhang, X.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', Li, P.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', and Li, H.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' Ambert: A pre-trained language  model with multi-grained tokenization.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' arXiv preprint  arXiv:2008.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='11869, 2020.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='  Zhu, J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', Zhu, X.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', Wang, W.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', Wang, X.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', Li, H.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', Wang, X.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=',  and Dai, J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' Uni-perceiver-moe: Learning sparse gen-  eralist models with conditional moes.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' arXiv preprint  arXiv:2206.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='04674, 2022.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='  Zhu, X.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', Zhu, J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', Li, H.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', Wu, X.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', Wang, X.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', Li, H.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', Wang,  X.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', and Dai, J.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' Uni-perceiver: Pre-training unified archi-  tecture for generic perception for zero-shot and few-shot  tasks.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' arXiv preprint arXiv:2112.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='01522, 2021.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='  Toward Building General Foundation Models for Language, Vision, and Vision-Language Understanding Tasks  A.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' Comparison of Recent Foundation Models  Table 8 shows an extensive comparison of recent foundation  models and X-FM on multiple axes.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' Previous work either (i)  perform best on uni-modal tasks (Liu et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', 2019;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' Peng et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=',  2022) or vision-language tasks (Zeng et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', 2021;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' 2022);' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' (2)  target a specific uni-modal domain along with part of vision-  and-language tasks (Wang et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', 2021a;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' Radford et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', 2021;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='  Jia et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', 2021;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' Wang et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', 2021c;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' Yu et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', 2022;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' Wang  et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', 2022b;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' Diao et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', 2022);' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' or (3) target all domains  but cannot perform best on all the tasks (Li et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', 2021c;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='  Singh et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', 2021;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' Zhu et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', 2022).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' Our model, X-FM, is  a general foundation model that can perform the best for  all the understanding tasks of language, vision, and vision  language.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='  B.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' Details of Downstream Tasks  Language Understanding.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='  We conduct experiments on GLUE benchmark including  MNLI (Williams et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', 2018), CoLA (Warstadt et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', 2019),  MRPC (Dolan & Brockett, 2005), QQP (Iyer et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', 2017),  SST-2 (Socher et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', 2013), QNLI (Rajpurkar et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', 2016),  RTE (Dagan et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', 2005;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' Haim et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', 2006;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' Giampiccolo  et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', 2007;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' Bentivogli et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', 2009), and STS-B (Agirre  et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', 2007).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' We follow the practice of BERT (Devlin et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=',  2019;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' Liu et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', 2019) and feed the input into the language  encoder, and the hidden state of the [CLS] is fed into a  new multi-class linear classifier or regression head.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='  Vision Understanding.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='  We conduct vision experiments on both fine-tuning and lin-  ear evaluation (linear eval).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' The linear evaluation follows  a common practice (Caron et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', 2021;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' He et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', 2020;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='  Singh et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', 2021) in self-supervised learning to evaluate  the representation quality, where the pre-trained backbone  model is frozen, and an MLP head is appended on top of it.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='  We choose 7 popular datasets following OmnVL (Wang  et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', 2022a):  ImageNet (Russakovsky et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', 2015),  Food101 (Bossard et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', 2014), CIFAR10 (Krizhevsky et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=',  2009), CIFAR100 (Krizhevsky et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', 2009), DTD (Cimpoi  et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', 2014), Pets (Parkhi et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', 2012) and Flowers102 (Nils-  back & Zisserman, 2008).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='  Vision-Language Understanding.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='  Image-Text Retrieval We evaluate X-FM on both  MSCOCO and Flickr30K datasets.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' We adopt the widely  used Karpathy split (Karpathy & Li, 2015) for both datasets.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='  Following the previous work (Li et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', 2021a;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' Zeng et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=',  2021;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' 2022), we first encode images and texts separately  and calculate s(I, T) to obtain the top-k candidates, and  then use the fusion encoder to re-rank the candidates.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='  Visual Question Answering The task requires the model  to predict an answer given an image and a question.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' We  evaluate X-FM on the VQA v2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='0 dataset (Goyal et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', 2017).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='  Following the previous work (Zeng et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', 2021), we use  a Transformer decoder to generate answers based on the  outputs of the fusion module.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' The decoder network shares  the same network architecture with the fusion encoder.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' Note  that we use an image resolution of 768*768 for the final  result of X-FMbase, and use an image resolution of 480*480  for X-FMlarge and X-FMbase in ablation studies for efficient  fine-tuning.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='  Visual Reasoning We evaluate X-FM on a widely used  benchmark NLVR2 (Suhr et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', 2019a).' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' The task allows the  model to determine whether a text describes the relations  between two images.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' Following previous work (Wang et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=',  2021a;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' Bao et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', 2022), we formulate the triplet input into  two image-text pairs, each containing the text description  and an image.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' We then concatenate the final output [CLS]  features of the fusion module of the two pairs to predict the  label.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='  C.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' Details of hyper parameters  Pre-training  X-FMbase is implemented with a 12-layer  language encoder, a 12-layer vision encoder, and a 12-layer  fusion encoder, 768 dimensions for hidden states, 3072 for  intermediate size, and 128 for maximum input length.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' X-  FMlarge is implemented with a 24-layer language encoder, a  24-layer vision encoder, and a 12-layer fusion encoder, 1024  dimensions for hidden states, 4096 for intermediate size, and  128 for maximum input length.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' We initialize the language  encoder with RoBERTa and the vision encoder with BEiTv2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='  The weight decay is set to 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='01 with β1 = 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='9, β2 = 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='98.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='  The learning rate is 1e-4 with a warm-up period for the first  2500 steps and then linearly decayed to 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' In each batch,  there are 3072 image-text pairs, 3072 images, and 8192 text-  only sentences.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' We use center-crop to resize each image  to the size of 224×224.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' The default settings are shown in  Table 9.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='  Fine-tuning  The learning rate is ∈ {1e-5, 2e-5, 5e-5} and  our model is optimized by AdamW.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' Because the image  resolution differs between pre-training and fine-tuning, the  position parameters are adapted using linear interpolation.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='  For all downstream tasks, we apply random resize crops  and horizontal flips augmentation during training.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' The de-  fault settings for text classification, image classification and  vision-language understanding are shown in Tables 10, 11,  12 and 13, respectively.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' Note that the resolution for VQA is  different as described in Section B.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='  Toward Building General Foundation Models for Language, Vision, and Vision-Language Understanding Tasks  Methods  Multimodal data  Pretraining Objectives  Fusion Arch.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='  Target Modalities  public  dataset(s)  size  Contr.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='  ITM  BBP  (M/P)LM  Unimodal  ST  CT  MT  V  CV&L  MV&L  L  RoBERTa (Liu et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', 2019)  –  –  –  –  –  –  –  MLM  –  –  –  –  –  –  �  BEiTv2 (Peng et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', 2022)  –  –  –  –  –  –  –  MIM  –  –  –  �  –  –  –  X-VLM (Zeng et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', 2021;' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' 2022)  �  Combination  5M  �  �  �  MLM  –  –  �  –  –  �  �  –  VLMo (Wang et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', 2021a)  �  Combination  5M  �  �  –  MLM  MLM+MIM  –  –  �  –  �  �  –  CLIP (Radford et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', 2021)  �  WebImageText  400M  �  –  –  –  –  –  –  –  �  �  –  –  ALIGN (Jia et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', 2021)  �  JFT  1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='8B  �  –  –  –  –  –  –  –  �  �  –  –  SimVLM (Wang et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', 2021c)  �  JFT  1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='8B  –  –  –  PrefixLM  PrefixLM  �  –  –  ∗  –  �  �  CoCa (Yu et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', 2022)  �  JFT  4.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='8B  �  –  –  LM  –  �  –  –  �  �  �  –  UNIMO-2 (Li et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', 2021c)  �  Combination  5M  –  �  –  MLM  VCL  �  –  –  �  �  �  �  OFA (Wang et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', 2022b)  �  Combination  15M  –  –  –  LM  LM  �  –  –  ∗  –  �  �  DaVinci (Diao et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', 2022)  �  Combination  46M  –  –  –  PrefixLM + PrefixIM  PrefixLM  �  –  –  �  –  �  �  FLAVA (Singh et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', 2021)  �  Combination  70M  �  �  –  MLM  MLM+MIM  �  –  –  �  �  �  �  Uni-Perceiver-MoE (Zhu et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', 2022)  �  Combination  116M  –  �  –  LM+MLM  LM+MLM+Classify.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='  �  –  –  �  �  �  �  X-FM  �  Combination  5M  �  �  �  MLM+MIM  MLM+MIM  –  �  –  �  �  �  �  Super-Large Models  Flamingo (Alayrac et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', 2022)  �  Combination  2.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='2B  –  –  –  LM  –  �  –  –  –  �  �  –  BEiT-v3 (Wang et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', 2022d)  �  Combination  21M  –  –  –  MLM  MLM+MIM  –  –  �  ∗  �  �  –  PaLI (Chen et al.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=', 2022)  �  WebImageText  41B  –  –  –  LM  –  �  –  –  �  �  �  �  Table 8: Comparison of recent foundation models in different modalities.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' Contr.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' indicates contrastive learning.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' ITM is  short for image-text matching.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' BBP represents boundary box prediction.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' (M/P)LM means image-conditioned (masked/prefix)  language modeling.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' V, CV&L, MV&L and L stand for vision tasks, cross-modal retrieval tasks, multi-modal fusion tasks  and language tasks respectively.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' ST, CT and MT are abbreviations for single Transformer, cross-attention Transformer and  multiway Transformer.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' VCL stands for visual contrastive learning.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' ∗ means the modality is partially targeted (SimVLM  and OFA include ImageNet.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=').' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' Giant models with over 1B parameters (e.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='g.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content=' BEiT-3) are in grey since they are not directly  comparable with other models.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='  config  value  optimizer  AdamW  learning rate  1e-4  weight decay  0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='01  optimizer momentum  β1, β2=0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='9, 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='999  language batch size  8192  vision batch size  3072  vision-language batch size  3072  learning rate schedule  linear decay  warmup steps  2500  training steps  200k  augmentation  RandomResizedCrop  image res  224*224  patch size  16  text length for MLM  128  text length for IMLM  30  Table 9: Pre-training setting.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='  config  value  optimizer  AdamW  learning rate  {1e-5, 2e-5, 5e-5}  weight decay  0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='0  optimizer momentum  β1, β2=0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='9, 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='999  batch size  {16, 32, 64}  learning rate schedule  linear decay  warmup ratio  0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='0  training epochs  {5, 10, 20}  Table 10: Text classification: GLUE setting.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='  config  value  optimizer  AdamW  learning rate  [2e-5, 4e-5]  weight decay  0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='01  optimizer momentum  β1, β2=0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='9, 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='999  batch size  [256, 2048]  learning rate schedule  linear decay  warmup rate  0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='1  training epochs  100  augmentation  RandomResizedCrop  image res  224*224  patch size  16  Table 11: Image classification: Linear probing setting.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='  Toward Building General Foundation Models for Language, Vision, and Vision-Language Understanding Tasks  config  value  optimizer  AdamW  learning rate  4e-5  minimal learning rate  1e-7  weight decay  0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='01  optimizer momentum  β1, β2=0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='9, 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='999  batch size  1024  learning rate schedule  linear decay  warmup rate  0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='1  training epochs  100  augmentation  RandomResizedCrop  image res  224*224  patch size  16  label smoothing  0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='1  mixup prob.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='  1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='0  cutmix prob.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='  1.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='0  Table 12: ImageNet classification: Fine-tuning setting.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='  config  value  optimizer  AdamW  learning rate  {1e-5, 2e-5, 5e-5}  weight decay  0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='01  optimizer momentum  β1, β2=0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='9, 0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='999  batch size  {64, 192, 512}  learning rate schedule  linear decay  warmup rate  0.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}
+page_content='1  training epochs  {10, 15, 20}  augmentation  RandomResizedCrop  image res  384*384  patch size  16  Table 13: Vision-Language understanding: fine-tuning set-  ting.' metadata={'source': '/home/zjlab/wf/langchain-ChatGLM/knowledge_base/2dE4T4oBgHgl3EQfagyV/content/2301.05065v1.pdf'}