date
timestamp[ns]date 2023-05-05 00:00:00
2025-06-25 00:00:00
| arxiv_id
stringlengths 10
10
| title
stringlengths 8
202
| authors
sequencelengths 1
942
| github
stringlengths 0
116
| abstract
stringlengths 165
1.92k
|
|---|---|---|---|---|---|
2025-06-04T00:00:00 | 2506.03119 | Controllable Human-centric Keyframe Interpolation with Generative Prior | [
"Zujin Guo",
"Size Wu",
"Zhongang Cai",
"Wei Li",
"Chen Change Loy"
] | Existing interpolation methods use pre-trained video diffusion priors to generate intermediate frames between sparsely sampled keyframes. In the absence of 3D geometric guidance, these methods struggle to produce plausible results for complex, articulated human motions and offer limited control over the synthesized dynamics. In this paper, we introduce PoseFuse3D Keyframe Interpolator (PoseFuse3D-KI), a novel framework that integrates 3D human guidance signals into the diffusion process for Controllable Human-centric Keyframe Interpolation (CHKI). To provide rich spatial and structural cues for interpolation, our PoseFuse3D, a 3D-informed control model, features a novel SMPL-X encoder that transforms 3D geometry and shape into the 2D latent conditioning space, alongside a fusion network that integrates these 3D cues with 2D pose embeddings. For evaluation, we build CHKI-Video, a new dataset annotated with both 2D poses and 3D SMPL-X parameters. We show that PoseFuse3D-KI consistently outperforms state-of-the-art baselines on CHKI-Video, achieving a 9% improvement in PSNR and a 38% reduction in LPIPS. Comprehensive ablations demonstrate that our PoseFuse3D model improves interpolation fidelity. |
|
2025-06-04T00:00:00 | 2506.00227 | Ctrl-Crash: Controllable Diffusion for Realistic Car Crashes | [
"Anthony Gosselin",
"Ge Ya Luo",
"Luis Lara",
"Florian Golemo",
"Derek Nowrouzezahrai",
"Liam Paull",
"Alexia Jolicoeur-Martineau",
"Christopher Pal"
] | Video diffusion techniques have advanced significantly in recent years; however, they struggle to generate realistic imagery of car crashes due to the scarcity of accident events in most driving datasets. Improving traffic safety requires realistic and controllable accident simulations. To tackle the problem, we propose Ctrl-Crash, a controllable car crash video generation model that conditions on signals such as bounding boxes, crash types, and an initial image frame. Our approach enables counterfactual scenario generation where minor variations in input can lead to dramatically different crash outcomes. To support fine-grained control at inference time, we leverage classifier-free guidance with independently tunable scales for each conditioning signal. Ctrl-Crash achieves state-of-the-art performance across quantitative video quality metrics (e.g., FVD and JEDi) and qualitative measurements based on a human-evaluation of physical realism and video quality compared to prior diffusion-based methods. |
|
2025-06-04T00:00:00 | 2506.01004 | Motion-Aware Concept Alignment for Consistent Video Editing | [
"Tong Zhang",
"Juan C Leon Alcazar",
"Bernard Ghanem"
] | We introduce MoCA-Video (Motion-Aware Concept Alignment in Video), a training-free framework bridging the gap between image-domain semantic mixing and video. Given a generated video and a user-provided reference image, MoCA-Video injects the semantic features of the reference image into a specific object within the video, while preserving the original motion and visual context. Our approach leverages a diagonal denoising schedule and class-agnostic segmentation to detect and track objects in the latent space and precisely control the spatial location of the blended objects. To ensure temporal coherence, we incorporate momentum-based semantic corrections and gamma residual noise stabilization for smooth frame transitions. We evaluate MoCA's performance using the standard SSIM, image-level LPIPS, temporal LPIPS, and introduce a novel metric CASS (Conceptual Alignment Shift Score) to evaluate the consistency and effectiveness of the visual shifts between the source prompt and the modified video frames. Using self-constructed dataset, MoCA-Video outperforms current baselines, achieving superior spatial consistency, coherent motion, and a significantly higher CASS score, despite having no training or fine-tuning. MoCA-Video demonstrates that structured manipulation in the diffusion noise trajectory allows for controllable, high-quality video synthesis. |
|
2025-06-04T00:00:00 | 2505.22704 | Training Language Models to Generate Quality Code with Program Analysis Feedback | [
"Feng Yao",
"Zilong Wang",
"Liyuan Liu",
"Junxia Cui",
"Li Zhong",
"Xiaohan Fu",
"Haohui Mai",
"Vish Krishnan",
"Jianfeng Gao",
"Jingbo Shang"
] | Code generation with large language models (LLMs), often termed vibe coding, is increasingly adopted in production but fails to ensure code quality, particularly in security (e.g., SQL injection vulnerabilities) and maintainability (e.g., missing type annotations). Existing methods, such as supervised fine-tuning and rule-based post-processing, rely on labor-intensive annotations or brittle heuristics, limiting their scalability and effectiveness. We propose REAL, a reinforcement learning framework that incentivizes LLMs to generate production-quality code using program analysis-guided feedback. Specifically, REAL integrates two automated signals: (1) program analysis detecting security or maintainability defects and (2) unit tests ensuring functional correctness. Unlike prior work, our framework is prompt-agnostic and reference-free, enabling scalable supervision without manual intervention. Experiments across multiple datasets and model scales demonstrate that REAL outperforms state-of-the-art methods in simultaneous assessments of functionality and code quality. Our work bridges the gap between rapid prototyping and production-ready code, enabling LLMs to deliver both speed and quality. |
|
2025-06-04T00:00:00 | 2506.02281 | Angles Don't Lie: Unlocking Training-Efficient RL Through the Model's Own Signals | [
"Qinsi Wang",
"Jinghan Ke",
"Hancheng Ye",
"Yueqian Lin",
"Yuzhe Fu",
"Jianyi Zhang",
"Kurt Keutzer",
"Chenfeng Xu",
"Yiran Chen"
] | https://github.com/wangqinsi1/GAINRL | Current Reinforcement Fine-tuning (RFT) paradigms for Large Language Models (LLMs) suffer from sample inefficiency due to the redundant exposure of identical queries under uniform data sampling. While previous work has explored curriculum learning via heuristic difficulty metrics, these strategies exhibit limitations by neglecting the intrinsic learning signals generated by the model itself, thus leading to suboptimal training regimes. In this paper, we identify a model-inherent signal termed angle concentration that effectively reflects an LLM's capacity to learn from specific data. We theoretically and empirically demonstrate a correlation between the angular distribution of token hidden state vectors and the resulting gradient, revealing a learning preference for data exhibiting higher angle concentration. Inspired by this finding, we propose GAIN-RL, a Gradient-driven Angle-Informed Navigated RL framework. By leveraging the model's intrinsic angle concentration signal, GAIN-RL dynamically selects training data in each epoch, ensuring consistently impactful gradient updates and thus significantly enhancing overall training efficiency. Empirical evaluations show that GAIN-RL (GRPO) achieves over a 2.5x acceleration in training efficiency across diverse mathematical and coding tasks and varying model scales. Furthermore, GAIN-RL (GRPO)'s efficient sampling yields data-efficient training, achieving better performance with half the original data compared to vanilla GRPO with full training data. Code is realsed at https://github.com/wangqinsi1/GAINRL/tree/main. |
2025-06-04T00:00:00 | 2506.03621 | Negative-Guided Subject Fidelity Optimization for Zero-Shot Subject-Driven Generation | [
"Chaehun Shin",
"Jooyoung Choi",
"Johan Barthelemy",
"Jungbeom Lee",
"Sungroh Yoon"
] | We present Subject Fidelity Optimization (SFO), a novel comparative learning framework for zero-shot subject-driven generation that enhances subject fidelity. Beyond supervised fine-tuning methods that rely only on positive targets and use the diffusion loss as in the pre-training stage, SFO introduces synthetic negative targets and explicitly guides the model to favor positives over negatives through pairwise comparison. For negative targets, we propose Condition-Degradation Negative Sampling (CDNS), which automatically generates distinctive and informative negatives by intentionally degrading visual and textual cues without expensive human annotations. Moreover, we reweight the diffusion timesteps to focus finetuning on intermediate steps where subject details emerge. Extensive experiments demonstrate that SFO with CDNS significantly outperforms baselines in terms of both subject fidelity and text alignment on a subject-driven generation benchmark. Project page: https://subjectfidelityoptimization.github.io/ |
|
2025-06-04T00:00:00 | 2506.03107 | ByteMorph: Benchmarking Instruction-Guided Image Editing with Non-Rigid Motions | [
"Di Chang",
"Mingdeng Cao",
"Yichun Shi",
"Bo Liu",
"Shengqu Cai",
"Shijie Zhou",
"Weilin Huang",
"Gordon Wetzstein",
"Mohammad Soleymani",
"Peng Wang"
] | Editing images with instructions to reflect non-rigid motions, camera viewpoint shifts, object deformations, human articulations, and complex interactions, poses a challenging yet underexplored problem in computer vision. Existing approaches and datasets predominantly focus on static scenes or rigid transformations, limiting their capacity to handle expressive edits involving dynamic motion. To address this gap, we introduce ByteMorph, a comprehensive framework for instruction-based image editing with an emphasis on non-rigid motions. ByteMorph comprises a large-scale dataset, ByteMorph-6M, and a strong baseline model built upon the Diffusion Transformer (DiT), named ByteMorpher. ByteMorph-6M includes over 6 million high-resolution image editing pairs for training, along with a carefully curated evaluation benchmark ByteMorph-Bench. Both capture a wide variety of non-rigid motion types across diverse environments, human figures, and object categories. The dataset is constructed using motion-guided data generation, layered compositing techniques, and automated captioning to ensure diversity, realism, and semantic coherence. We further conduct a comprehensive evaluation of recent instruction-based image editing methods from both academic and commercial domains. |
|
2025-06-04T00:00:00 | 2506.00189 | Control-R: Towards controllable test-time scaling | [
"Di Zhang",
"Weida Wang",
"Junxian Li",
"Xunzhi Wang",
"Jiatong Li",
"Jianbo Wu",
"Jingdi Lei",
"Haonan He",
"Peng Ye",
"Shufei Zhang",
"Wanli Ouyang",
"Yuqiang Li",
"Dongzhan Zhou"
] | This paper target in addressing the challenges of underthinking and overthinking in long chain-of-thought (CoT) reasoning for Large Reasoning Models (LRMs) by introducing Reasoning Control Fields (RCF)--a novel test-time approach that injects structured control signals to guide reasoning from a tree search perspective. RCF enables models to adjust reasoning effort according to given control conditions when solving complex tasks. Additionally, we present the Control-R-4K dataset, which consists of challenging problems annotated with detailed reasoning processes and corresponding control fields. To further enhance reasoning control, we propose a Conditional Distillation Finetuning (CDF) method, which trains model--particularly Control-R-32B--to effectively adjust reasoning effort during test time. Experimental results on benchmarks such as AIME2024 and MATH500 demonstrate that our approach achieves state-of-the-art performance at the 32B scale while enabling a controllable Long CoT reasoning process (L-CoT). Overall, this work introduces an effective paradigm for controllable test-time scaling reasoning. |
|
2025-06-05T00:00:00 | 2506.04133 | TRiSM for Agentic AI: A Review of Trust, Risk, and Security Management in LLM-based Agentic Multi-Agent Systems | [
"Shaina Raza",
"Ranjan Sapkota",
"Manoj Karkee",
"Christos Emmanouilidis"
] | Agentic AI systems, built on large language models (LLMs) and deployed in multi-agent configurations, are redefining intelligent autonomy, collaboration and decision-making across enterprise and societal domains. This review presents a structured analysis of Trust, Risk, and Security Management (TRiSM) in the context of LLM-based agentic multi-agent systems (AMAS). We begin by examining the conceptual foundations of agentic AI, its architectural differences from traditional AI agents, and the emerging system designs that enable scalable, tool-using autonomy. The TRiSM in the agentic AI framework is then detailed through four pillars governance, explainability, ModelOps, and privacy/security each contextualized for agentic LLMs. We identify unique threat vectors and introduce a comprehensive risk taxonomy for the agentic AI applications, supported by case studies illustrating real-world vulnerabilities. Furthermore, the paper also surveys trust-building mechanisms, transparency and oversight techniques, and state-of-the-art explainability strategies in distributed LLM agent systems. Additionally, metrics for evaluating trust, interpretability, and human-centered performance are reviewed alongside open benchmarking challenges. Security and privacy are addressed through encryption, adversarial defense, and compliance with evolving AI regulations. The paper concludes with a roadmap for responsible agentic AI, proposing research directions to align emerging multi-agent systems with robust TRiSM principles for safe, accountable, and transparent deployment. |
|
2025-06-05T00:00:00 | 2506.02294 | Improving Knowledge Distillation Under Unknown Covariate Shift Through Confidence-Guided Data Augmentation | [
"Niclas Popp",
"Kevin Alexander Laube",
"Matthias Hein",
"Lukas Schott"
] | Large foundation models trained on extensive datasets demonstrate strong zero-shot capabilities in various domains. To replicate their success when data and model size are constrained, knowledge distillation has become an established tool for transferring knowledge from foundation models to small student networks. However, the effectiveness of distillation is critically limited by the available training data. This work addresses the common practical issue of covariate shift in knowledge distillation, where spurious features appear during training but not at test time. We ask the question: when these spurious features are unknown, yet a robust teacher is available, is it possible for a student to also become robust to them? We address this problem by introducing a novel diffusion-based data augmentation strategy that generates images by maximizing the disagreement between the teacher and the student, effectively creating challenging samples that the student struggles with. Experiments demonstrate that our approach significantly improves worst group and mean group accuracy on CelebA and SpuCo Birds as well as the spurious mAUC on spurious ImageNet under covariate shift, outperforming state-of-the-art diffusion-based data augmentation baselines |
|
2025-06-05T00:00:00 | 2506.03569 | MiMo-VL Technical Report | [
"Xiaomi LLM-Core Team",
"Zihao Yue",
"Zhenru Lin",
"Yifan Song",
"Weikun Wang",
"Shuhuai Ren",
"Shuhao Gu",
"Shicheng Li",
"Peidian Li",
"Liang Zhao",
"Lei Li",
"Kainan Bao",
"Hao Tian",
"Hailin Zhang",
"Gang Wang",
"Dawei Zhu",
"Cici",
"Chenhong He",
"Bowen Ye",
"Bowen Shen",
"Zihan Zhang",
"Zihan Jiang",
"Zhixian Zheng",
"Zhichao Song",
"Zhenbo Luo",
"Yue Yu",
"Yudong Wang",
"Yuanyuan Tian",
"Yu Tu",
"Yihan Yan",
"Yi Huang",
"Xu Wang",
"Xinzhe Xu",
"Xingchen Song",
"Xing Zhang",
"Xing Yong",
"Xin Zhang",
"Xiangwei Deng",
"Wenyu Yang",
"Wenhan Ma",
"Weiwei Lv",
"Weiji Zhuang",
"Wei Liu",
"Sirui Deng",
"Shuo Liu",
"Shimao Chen",
"Shihua Yu",
"Shaohui Liu",
"Shande Wang",
"Rui Ma",
"Qiantong Wang",
"Peng Wang",
"Nuo Chen",
"Menghang Zhu",
"Kangyang Zhou",
"Kang Zhou",
"Kai Fang",
"Jun Shi",
"Jinhao Dong",
"Jiebao Xiao",
"Jiaming Xu",
"Huaqiu Liu",
"Hongshen Xu",
"Heng Qu",
"Haochen Zhao",
"Hanglong Lv",
"Guoan Wang",
"Duo Zhang",
"Dong Zhang",
"Di Zhang",
"Chong Ma",
"Chang Liu",
"Can Cai",
"Bingquan Xia"
] | https://github.com/XiaomiMiMo/MiMo-VL | We open-source MiMo-VL-7B-SFT and MiMo-VL-7B-RL, two powerful vision-language models delivering state-of-the-art performance in both general visual understanding and multimodal reasoning. MiMo-VL-7B-RL outperforms Qwen2.5-VL-7B on 35 out of 40 evaluated tasks, and scores 59.4 on OlympiadBench, surpassing models with up to 78B parameters. For GUI grounding applications, it sets a new standard with 56.1 on OSWorld-G, even outperforming specialized models such as UI-TARS. Our training combines four-stage pre-training (2.4 trillion tokens) with Mixed On-policy Reinforcement Learning (MORL) integrating diverse reward signals. We identify the importance of incorporating high-quality reasoning data with long Chain-of-Thought into pre-training stages, and the benefits of mixed RL despite challenges in simultaneous multi-domain optimization. We also contribute a comprehensive evaluation suite covering 50+ tasks to promote reproducibility and advance the field. The model checkpoints and full evaluation suite are available at https://github.com/XiaomiMiMo/MiMo-VL. |
2025-06-05T00:00:00 | 2505.24500 | TimeHC-RL: Temporal-aware Hierarchical Cognitive Reinforcement Learning for Enhancing LLMs' Social Intelligence | [
"Guiyang Hou",
"Xing Gao",
"Yuchuan Wu",
"Xiang Huang",
"Wenqi Zhang",
"Zhe Zheng",
"Yongliang Shen",
"Jialu Du",
"Fei Huang",
"Yongbin Li",
"Weiming Lu"
] | Recently, Large Language Models (LLMs) have made significant progress in IQ-related domains that require careful thinking, such as mathematics and coding. However, enhancing LLMs' cognitive development in social domains, particularly from a post-training perspective, remains underexplored. Recognizing that the social world follows a distinct timeline and requires a richer blend of cognitive modes (from intuitive reactions (System 1) and surface-level thinking to deliberate thinking (System 2)) than mathematics, which primarily relies on System 2 cognition (careful, step-by-step reasoning), we introduce Temporal-aware Hierarchical Cognitive Reinforcement Learning (TimeHC-RL) for enhancing LLMs' social intelligence. In our experiments, we systematically explore improving LLMs' social intelligence and validate the effectiveness of the TimeHC-RL method, through five other post-training paradigms and two test-time intervention paradigms on eight datasets with diverse data patterns. Experimental results reveal the superiority of our proposed TimeHC-RL method compared to the widely adopted System 2 RL method. It gives the 7B backbone model wings, enabling it to rival the performance of advanced models like DeepSeek-R1 and OpenAI-O3. Additionally, the systematic exploration from post-training and test-time interventions perspectives to improve LLMs' social intelligence has uncovered several valuable insights. |
|
2025-06-05T00:00:00 | 2506.04180 | SuperWriter: Reflection-Driven Long-Form Generation with Large Language Models | [
"Yuhao Wu",
"Yushi Bai",
"Zhiqiang Hu",
"Juanzi Li",
"Roy Ka-Wei Lee"
] | Long-form text generation remains a significant challenge for large language models (LLMs), particularly in maintaining coherence, ensuring logical consistency, and preserving text quality as sequence length increases. To address these limitations, we propose SuperWriter-Agent, an agent-based framework designed to enhance the quality and consistency of long-form text generation. SuperWriter-Agent introduces explicit structured thinking-through planning and refinement stages into the generation pipeline, guiding the model to follow a more deliberate and cognitively grounded process akin to that of a professional writer. Based on this framework, we construct a supervised fine-tuning dataset to train a 7B SuperWriter-LM. We further develop a hierarchical Direct Preference Optimization (DPO) procedure that uses Monte Carlo Tree Search (MCTS) to propagate final quality assessments and optimize each generation step accordingly. Empirical results across diverse benchmarks demonstrate that SuperWriter-LM achieves state-of-the-art performance, surpassing even larger-scale baseline models in both automatic evaluation and human evaluation. Furthermore, comprehensive ablation studies demonstrate the effectiveness of hierarchical DPO and underscore the value of incorporating structured thinking steps to improve the quality of long-form text generation. |
|
2025-06-05T00:00:00 | 2506.04228 | LayerFlow: A Unified Model for Layer-aware Video Generation | [
"Sihui Ji",
"Hao Luo",
"Xi Chen",
"Yuanpeng Tu",
"Yiyang Wang",
"Hengshuang Zhao"
] | We present LayerFlow, a unified solution for layer-aware video generation. Given per-layer prompts, LayerFlow generates videos for the transparent foreground, clean background, and blended scene. It also supports versatile variants like decomposing a blended video or generating the background for the given foreground and vice versa. Starting from a text-to-video diffusion transformer, we organize the videos for different layers as sub-clips, and leverage layer embeddings to distinguish each clip and the corresponding layer-wise prompts. In this way, we seamlessly support the aforementioned variants in one unified framework. For the lack of high-quality layer-wise training videos, we design a multi-stage training strategy to accommodate static images with high-quality layer annotations. Specifically, we first train the model with low-quality video data. Then, we tune a motion LoRA to make the model compatible with static frames. Afterward, we train the content LoRA on the mixture of image data with high-quality layered images along with copy-pasted video data. During inference, we remove the motion LoRA thus generating smooth videos with desired layers. |
|
2025-06-05T00:00:00 | 2506.04158 | Image Editing As Programs with Diffusion Models | [
"Yujia Hu",
"Songhua Liu",
"Zhenxiong Tan",
"Xingyi Yang",
"Xinchao Wang"
] | https://github.com/YujiaHu1109/IEAP | While diffusion models have achieved remarkable success in text-to-image generation, they encounter significant challenges with instruction-driven image editing. Our research highlights a key challenge: these models particularly struggle with structurally inconsistent edits that involve substantial layout changes. To mitigate this gap, we introduce Image Editing As Programs (IEAP), a unified image editing framework built upon the Diffusion Transformer (DiT) architecture. At its core, IEAP approaches instructional editing through a reductionist lens, decomposing complex editing instructions into sequences of atomic operations. Each operation is implemented via a lightweight adapter sharing the same DiT backbone and is specialized for a specific type of edit. Programmed by a vision-language model (VLM)-based agent, these operations collaboratively support arbitrary and structurally inconsistent transformations. By modularizing and sequencing edits in this way, IEAP generalizes robustly across a wide range of editing tasks, from simple adjustments to substantial structural changes. Extensive experiments demonstrate that IEAP significantly outperforms state-of-the-art methods on standard benchmarks across various editing scenarios. In these evaluations, our framework delivers superior accuracy and semantic fidelity, particularly for complex, multi-step instructions. Codes are available at https://github.com/YujiaHu1109/IEAP. |
2025-06-05T00:00:00 | 2506.04108 | Rectified Sparse Attention | [
"Yutao Sun",
"Tianzhu Ye",
"Li Dong",
"Yuqing Xia",
"Jian Chen",
"Yizhao Gao",
"Shijie Cao",
"Jianyong Wang",
"Furu Wei"
] | Efficient long-sequence generation is a critical challenge for Large Language Models. While recent sparse decoding methods improve efficiency, they suffer from KV cache misalignment, where approximation errors accumulate and degrade generation quality. In this work, we propose Rectified Sparse Attention (ReSA), a simple yet effective method that combines block-sparse attention with periodic dense rectification. By refreshing the KV cache at fixed intervals using a dense forward pass, ReSA bounds error accumulation and preserves alignment with the pretraining distribution. Experiments across math reasoning, language modeling, and retrieval tasks demonstrate that ReSA achieves near-lossless generation quality with significantly improved efficiency. Notably, ReSA delivers up to 2.42times end-to-end speedup under decoding at 256K sequence length, making it a practical solution for scalable long-context inference. Code is available at https://aka.ms/ReSA-LM. |
|
2025-06-05T00:00:00 | 2506.04034 | Rex-Thinker: Grounded Object Referring via Chain-of-Thought Reasoning | [
"Qing Jiang",
"Xingyu Chen",
"Zhaoyang Zeng",
"Junzhi Yu",
"Lei Zhang"
] | Object referring aims to detect all objects in an image that match a given natural language description. We argue that a robust object referring model should be grounded, meaning its predictions should be both explainable and faithful to the visual content. Specifically, it should satisfy two key properties: 1) Verifiable, by producing interpretable reasoning that justifies its predictions and clearly links them to visual evidence; and 2) Trustworthy, by learning to abstain when no object in the image satisfies the given expression. However, most methods treat referring as a direct bounding box prediction task, offering limited interpretability and struggling to reject expressions with no matching object. In this work, we propose Rex-Thinker, a model that formulates object referring as an explicit CoT reasoning task. Given a referring expression, we first identify all candidate object instances corresponding to the referred object category. Rex-Thinker then performs step-by-step reasoning over each candidate to assess whether it matches the given expression, before making a final prediction. To support this paradigm, we construct a large-scale CoT-style referring dataset named HumanRef-CoT by prompting GPT-4o on the HumanRef dataset. Each reasoning trace follows a structured planning, action, and summarization format, enabling the model to learn decomposed, interpretable reasoning over object candidates. We then train Rex-Thinker in two stages: a cold-start supervised fine-tuning phase to teach the model how to perform structured reasoning, followed by GRPO-based RL learning to improve accuracy and generalization. Experiments show that our approach outperforms standard baselines in both precision and interpretability on in-domain evaluation, while also demonstrating improved ability to reject hallucinated outputs and strong generalization in out-of-domain settings. |
|
2025-06-05T00:00:00 | 2506.03448 | RefEdit: A Benchmark and Method for Improving Instruction-based Image Editing Model on Referring Expressions | [
"Bimsara Pathiraja",
"Maitreya Patel",
"Shivam Singh",
"Yezhou Yang",
"Chitta Baral"
] | Despite recent advances in inversion and instruction-based image editing, existing approaches primarily excel at editing single, prominent objects but significantly struggle when applied to complex scenes containing multiple entities. To quantify this gap, we first introduce RefEdit-Bench, a rigorous real-world benchmark rooted in RefCOCO, where even baselines trained on millions of samples perform poorly. To overcome this limitation, we introduce RefEdit -- an instruction-based editing model trained on our scalable synthetic data generation pipeline. Our RefEdit, trained on only 20,000 editing triplets, outperforms the Flux/SD3 model-based baselines trained on millions of data. Extensive evaluations across various benchmarks demonstrate that our model not only excels in referring expression tasks but also enhances performance on traditional benchmarks, achieving state-of-the-art results comparable to closed-source methods. We release data \& checkpoint for reproducibility. |
|
2025-06-05T00:00:00 | 2506.03106 | Critique-GRPO: Advancing LLM Reasoning with Natural Language and Numerical Feedback | [
"Xiaoying Zhang",
"Hao Sun",
"Yipeng Zhang",
"Kaituo Feng",
"Chaochao Lu",
"Chao Yang",
"Helen Meng"
] | Recent advances in reinforcement learning (RL) with numerical feedback, such as scalar rewards, have significantly enhanced the complex reasoning capabilities of large language models (LLMs). Despite this success, we identify three key challenges encountered by RL with solely numerical feedback: performance plateaus, limited effectiveness of self-reflection, and persistent failures. We then demonstrate that RL-finetuned models, even after exhibiting performance plateaus, can generate correct refinements on persistently failed problems by leveraging natural language feedback in the form of critiques. Building on this insight, we propose Critique-GRPO, an online RL framework that integrates both natural language and numerical feedback for effective policy optimization. Critique-GRPO enables LLMs to learn from initial responses and critique-guided refinements simultaneously while maintaining exploration. Extensive experiments using Qwen2.5-7B-Base and Qwen3-8B-Base show that Critique-GRPO consistently outperforms supervised learning-based and RL-based fine-tuning approaches across eight challenging mathematical, STEM, and general reasoning tasks, improving average pass@1 scores by approximately 4.5% and 5%, respectively. Notably, Critique-GRPO surpasses a strong baseline that incorporates expert demonstrations within online RL. Further analysis reveals two critical insights about policy exploration: (1) higher entropy does not always guarantee efficient learning from exploration, and (2) longer responses do not necessarily lead to more effective exploration. |
|
2025-06-05T00:00:00 | 2506.02945 | Quantitative LLM Judges | [
"Aishwarya Sahoo",
"Jeevana Kruthi Karnuthala",
"Tushar Parmanand Budhwani",
"Pranchal Agarwal",
"Sankaran Vaidyanathan",
"Alexa Siu",
"Franck Dernoncourt",
"Jennifer Healey",
"Nedim Lipka",
"Ryan Rossi",
"Uttaran Bhattacharya",
"Branislav Kveton"
] | LLM-as-a-judge is a framework in which a large language model (LLM) automatically evaluates the output of another LLM. We propose quantitative LLM judges, which align evaluation scores of existing LLM judges to human scores in a given domain using regression models. The models are trained to improve the score of the original judge by using the judge's textual evaluation and score. We present four quantitative judges for different types of absolute and relative feedback, which showcases the generality and versatility of our framework. Our framework is more computationally efficient than supervised fine-tuning and can be more statistically efficient when human feedback is limited, which is expected in most applications of our work. We validate these claims empirically on four datasets using two base judges. Our experiments show that quantitative judges can effectively improve the predictive power of existing judges through post-hoc modeling. |
|
2025-06-05T00:00:00 | 2506.01344 | Follow the Flow: Fine-grained Flowchart Attribution with Neurosymbolic Agents | [
"Manan Suri",
"Puneet Mathur",
"Nedim Lipka",
"Franck Dernoncourt",
"Ryan A. Rossi",
"Vivek Gupta",
"Dinesh Manocha"
] | Flowcharts are a critical tool for visualizing decision-making processes. However, their non-linear structure and complex visual-textual relationships make it challenging to interpret them using LLMs, as vision-language models frequently hallucinate nonexistent connections and decision paths when analyzing these diagrams. This leads to compromised reliability for automated flowchart processing in critical domains such as logistics, health, and engineering. We introduce the task of Fine-grained Flowchart Attribution, which traces specific components grounding a flowchart referring LLM response. Flowchart Attribution ensures the verifiability of LLM predictions and improves explainability by linking generated responses to the flowchart's structure. We propose FlowPathAgent, a neurosymbolic agent that performs fine-grained post hoc attribution through graph-based reasoning. It first segments the flowchart, then converts it into a structured symbolic graph, and then employs an agentic approach to dynamically interact with the graph, to generate attribution paths. Additionally, we present FlowExplainBench, a novel benchmark for evaluating flowchart attributions across diverse styles, domains, and question types. Experimental results show that FlowPathAgent mitigates visual hallucinations in LLM answers over flowchart QA, outperforming strong baselines by 10-14% on our proposed FlowExplainBench dataset. |
|
2025-06-05T00:00:00 | 2505.23807 | DLP: Dynamic Layerwise Pruning in Large Language Models | [
"Yuli Chen",
"Bo Cheng",
"Jiale Han",
"Yingying Zhang",
"Yingting Li",
"Shuhao Zhang"
] | https://github.com/ironartisan/DLP | Pruning has recently been widely adopted to reduce the parameter scale and improve the inference efficiency of Large Language Models (LLMs). Mainstream pruning techniques often rely on uniform layerwise pruning strategies, which can lead to severe performance degradation at high sparsity levels. Recognizing the varying contributions of different layers in LLMs, recent studies have shifted their focus toward non-uniform layerwise pruning. However, these approaches often rely on pre-defined values, which can result in suboptimal performance. To overcome these limitations, we propose a novel method called Dynamic Layerwise Pruning (DLP). This approach adaptively determines the relative importance of each layer by integrating model weights with input activation information, assigning pruning rates accordingly. Experimental results show that DLP effectively preserves model performance at high sparsity levels across multiple LLMs. Specifically, at 70% sparsity, DLP reduces the perplexity of LLaMA2-7B by 7.79 and improves the average accuracy by 2.7% compared to state-of-the-art methods. Moreover, DLP is compatible with various existing LLM compression techniques and can be seamlessly integrated into Parameter-Efficient Fine-Tuning (PEFT). We release the code at https://github.com/ironartisan/DLP to facilitate future research. |
2025-06-05T00:00:00 | 2505.21541 | DiffDecompose: Layer-Wise Decomposition of Alpha-Composited Images via Diffusion Transformers | [
"Zitong Wang",
"Hang Zhao",
"Qianyu Zhou",
"Xuequan Lu",
"Xiangtai Li",
"Yiren Song"
] | https://github.com/Wangzt1121/DiffDecompose | Diffusion models have recently motivated great success in many generation tasks like object removal. Nevertheless, existing image decomposition methods struggle to disentangle semi-transparent or transparent layer occlusions due to mask prior dependencies, static object assumptions, and the lack of datasets. In this paper, we delve into a novel task: Layer-Wise Decomposition of Alpha-Composited Images, aiming to recover constituent layers from single overlapped images under the condition of semi-transparent/transparent alpha layer non-linear occlusion. To address challenges in layer ambiguity, generalization, and data scarcity, we first introduce AlphaBlend, the first large-scale and high-quality dataset for transparent and semi-transparent layer decomposition, supporting six real-world subtasks (e.g., translucent flare removal, semi-transparent cell decomposition, glassware decomposition). Building on this dataset, we present DiffDecompose, a diffusion Transformer-based framework that learns the posterior over possible layer decompositions conditioned on the input image, semantic prompts, and blending type. Rather than regressing alpha mattes directly, DiffDecompose performs In-Context Decomposition, enabling the model to predict one or multiple layers without per-layer supervision, and introduces Layer Position Encoding Cloning to maintain pixel-level correspondence across layers. Extensive experiments on the proposed AlphaBlend dataset and public LOGO dataset verify the effectiveness of DiffDecompose. The code and dataset will be available upon paper acceptance. Our code will be available at: https://github.com/Wangzt1121/DiffDecompose. |
2025-06-05T00:00:00 | 2506.02921 | A Controllable Examination for Long-Context Language Models | [
"Yijun Yang",
"Zeyu Huang",
"Wenhao Zhu",
"Zihan Qiu",
"Fei Yuan",
"Jeff Z. Pan",
"Ivan Titov"
] | Existing frameworks for evaluating long-context language models (LCLM) can be broadly categorized into real-world and synthetic tasks. Despite their utility, both approaches are accompanied by certain intrinsic limitations. Real-world tasks are too complex to interpret or characterize and are susceptible to data contamination. In contrast, synthetic tasks often adopt the needle-in-the-haystack (NIAH) format, wherein a lack of coherence between the "needle" and the "haystack" compromises their validity as proxies for realistic applications. In response to these challenges, we posit that an ideal long-context evaluation framework should be characterized by three essential features: seamless context, controllable setting, and sound evaluation. This study introduces LongBioBench, a novel benchmark that utilizes artificially generated biographies as a controlled environment for assessing LCLMs across dimensions of understanding, reasoning, and trustworthiness. Our experimental evaluation, which includes 18 LCLMs in total, demonstrates that most models still exhibit deficiencies in semantic understanding and elementary reasoning over retrieved results and are less trustworthy as context length increases. Our further analysis indicates some design choices employed by existing synthetic benchmarks, such as contextual non-coherence, numerical needles, and the absence of distractors, rendering them vulnerable to test the model long-context capabilities. Moreover, we also reveal that long-context continual pretraining primarily adjusts RoPE embedding to accommodate extended context lengths. To sum up, compared to previous synthetic benchmarks, LongBioBench achieves a better trade-off between mirroring authentic language tasks and maintaining controllability, and is highly interpretable and configurable. |
|
2025-06-05T00:00:00 | 2506.02592 | Beyond the Surface: Measuring Self-Preference in LLM Judgments | [
"Zhi-Yuan Chen",
"Hao Wang",
"Xinyu Zhang",
"Enrui Hu",
"Yankai Lin"
] | https://github.com/zhiyuanc2001/self-preference | Recent studies show that large language models (LLMs) exhibit self-preference bias when serving as judges, meaning they tend to favor their own responses over those generated by other models. Existing methods typically measure this bias by calculating the difference between the scores a judge model assigns to its own responses and those it assigns to responses from other models. However, this approach conflates self-preference bias with response quality, as higher-quality responses from the judge model may also lead to positive score differences, even in the absence of bias. To address this issue, we introduce gold judgments as proxies for the actual quality of responses and propose the DBG score, which measures self-preference bias as the difference between the scores assigned by the judge model to its own responses and the corresponding gold judgments. Since gold judgments reflect true response quality, the DBG score mitigates the confounding effect of response quality on bias measurement. Using the DBG score, we conduct comprehensive experiments to assess self-preference bias across LLMs of varying versions, sizes, and reasoning abilities. Additionally, we investigate two factors that influence and help alleviate self-preference bias: response text style and the post-training data of judge models. Finally, we explore potential underlying mechanisms of self-preference bias from an attention-based perspective. Our code and data are available at https://github.com/zhiyuanc2001/self-preference. |
2025-06-05T00:00:00 | 2506.03099 | TalkingMachines: Real-Time Audio-Driven FaceTime-Style Video via Autoregressive Diffusion Models | [
"Chetwin Low",
"Weimin Wang"
] | In this paper, we present TalkingMachines -- an efficient framework that transforms pretrained video generation models into real-time, audio-driven character animators. TalkingMachines enables natural conversational experiences by integrating an audio large language model (LLM) with our video generation foundation model. Our primary contributions include: (1) We adapt a pretrained SOTA image-to-video DiT into an audio-driven avatar generation model of 18 billion parameters; (2) We enable infinite video streaming without error accumulation through asymmetric knowledge distillation from a bidirectional teacher model into a sparse causal, autoregressive student model; (3) We design a high-throughput, low-latency inference pipeline incorporating several key engineering optimizations such as: (a) disaggregation of the DiT and VAE decoder across separate devices, (b) efficient overlap of inter-device communication and computation using CUDA streams, (c) elimination of redundant recomputations to maximize frame-generation throughput. Please see demo videos here - https://aaxwaz.github.io/TalkingMachines/ |
|
2025-06-05T00:00:00 | 2506.03295 | Unleashing the Reasoning Potential of Pre-trained LLMs by Critique Fine-Tuning on One Problem | [
"Yubo Wang",
"Ping Nie",
"Kai Zou",
"Lijun Wu",
"Wenhu Chen"
] | We have witnessed that strong LLMs like Qwen-Math, MiMo, and Phi-4 possess immense reasoning potential inherited from the pre-training stage. With reinforcement learning (RL), these models can improve dramatically on reasoning tasks. Recent studies have shown that even RL on a single problem can unleash these models' reasoning capabilities. However, RL is not only expensive but also unstable. Even one-shot RL requires hundreds of GPU hours. This raises a critical question: Is there a more efficient way to unleash the reasoning potential of these powerful base LLMs? In this work, we demonstrate that Critique Fine-Tuning (CFT) on only one problem can effectively unleash the reasoning potential of LLMs. Our method constructs critique data by collecting diverse model-generated solutions to a single problem and using teacher LLMs to provide detailed critiques. We fine-tune Qwen and Llama family models, ranging from 1.5B to 14B parameters, on the CFT data and observe significant performance gains across diverse reasoning tasks. For example, with just 5 GPU hours of training, Qwen-Math-7B-CFT show an average improvement of 15% on six math benchmarks and 16% on three logic reasoning benchmarks. These results are comparable to or even surpass the results from RL with 20x less compute. Ablation studies reveal the robustness of one-shot CFT across different prompt problems. These results highlight one-shot CFT as a simple, general, and compute-efficient approach to unleashing the reasoning capabilities of modern LLMs. |
|
2025-06-05T00:00:00 | 2506.04207 | Advancing Multimodal Reasoning: From Optimized Cold Start to Staged Reinforcement Learning | [
"Shuang Chen",
"Yue Guo",
"Zhaochen Su",
"Yafu Li",
"Yulun Wu",
"Jiacheng Chen",
"Jiayu Chen",
"Weijie Wang",
"Xiaoye Qu",
"Yu Cheng"
] | Inspired by the remarkable reasoning capabilities of Deepseek-R1 in complex textual tasks, many works attempt to incentivize similar capabilities in Multimodal Large Language Models (MLLMs) by directly applying reinforcement learning (RL). However, they still struggle to activate complex reasoning. In this paper, rather than examining multimodal RL in isolation, we delve into current training pipelines and identify three crucial phenomena: 1) Effective cold start initialization is critical for enhancing MLLM reasoning. Intriguingly, we find that initializing with carefully selected text data alone can lead to performance surpassing many recent multimodal reasoning models, even before multimodal RL. 2) Standard GRPO applied to multimodal RL suffers from gradient stagnation, which degrades training stability and performance. 3) Subsequent text-only RL training, following the multimodal RL phase, further enhances multimodal reasoning. This staged training approach effectively balances perceptual grounding and cognitive reasoning development. By incorporating the above insights and addressing multimodal RL issues, we introduce ReVisual-R1, achieving a new state-of-the-art among open-source 7B MLLMs on challenging benchmarks including MathVerse, MathVision, WeMath, LogicVista, DynaMath, and challenging AIME2024 and AIME2025. |
|
2025-06-05T00:00:00 | 2506.01320 | Ψ-Sampler: Initial Particle Sampling for SMC-Based Inference-Time Reward Alignment in Score Models | [
"Taehoon Yoon",
"Yunhong Min",
"Kyeongmin Yeo",
"Minhyuk Sung"
] | We introduce Psi-Sampler, an SMC-based framework incorporating pCNL-based initial particle sampling for effective inference-time reward alignment with a score-based generative model. Inference-time reward alignment with score-based generative models has recently gained significant traction, following a broader paradigm shift from pre-training to post-training optimization. At the core of this trend is the application of Sequential Monte Carlo (SMC) to the denoising process. However, existing methods typically initialize particles from the Gaussian prior, which inadequately captures reward-relevant regions and results in reduced sampling efficiency. We demonstrate that initializing from the reward-aware posterior significantly improves alignment performance. To enable posterior sampling in high-dimensional latent spaces, we introduce the preconditioned Crank-Nicolson Langevin (pCNL) algorithm, which combines dimension-robust proposals with gradient-informed dynamics. This approach enables efficient and scalable posterior sampling and consistently improves performance across various reward alignment tasks, including layout-to-image generation, quantity-aware generation, and aesthetic-preference generation, as demonstrated in our experiments. |
|
2025-06-05T00:00:00 | 2506.03139 | SVGenius: Benchmarking LLMs in SVG Understanding, Editing and Generation | [
"Siqi Chen",
"Xinyu Dong",
"Haolei Xu",
"Xingyu Wu",
"Fei Tang",
"Hang Zhang",
"Yuchen Yan",
"Linjuan Wu",
"Wenqi Zhang",
"Guiyang Hou",
"Yongliang Shen",
"Weiming Lu",
"Yueting Zhuang"
] | Large Language Models (LLMs) and Multimodal LLMs have shown promising capabilities for SVG processing, yet existing benchmarks suffer from limited real-world coverage, lack of complexity stratification, and fragmented evaluation paradigms. We introduce SVGenius, a comprehensive benchmark comprising 2,377 queries across three progressive dimensions: understanding, editing, and generation. Built on real-world data from 24 application domains with systematic complexity stratification, SVGenius evaluates models through 8 task categories and 18 metrics. We assess 22 mainstream models spanning different scales, architectures, training paradigms, and accessibility levels. Our analysis reveals that while proprietary models significantly outperform open-source counterparts, all models exhibit systematic performance degradation with increasing complexity, indicating fundamental limitations in current approaches; however, reasoning-enhanced training proves more effective than pure scaling for overcoming these limitations, though style transfer remains the most challenging capability across all model types. SVGenius establishes the first systematic evaluation framework for SVG processing, providing crucial insights for developing more capable vector graphics models and advancing automated graphic design applications. Appendix and supplementary materials (including all data and code) are available at https://zju-real.github.io/SVGenius. |
|
2025-06-05T00:00:00 | 2506.03517 | DenseDPO: Fine-Grained Temporal Preference Optimization for Video Diffusion Models | [
"Ziyi Wu",
"Anil Kag",
"Ivan Skorokhodov",
"Willi Menapace",
"Ashkan Mirzaei",
"Igor Gilitschenski",
"Sergey Tulyakov",
"Aliaksandr Siarohin"
] | Direct Preference Optimization (DPO) has recently been applied as a post-training technique for text-to-video diffusion models. To obtain training data, annotators are asked to provide preferences between two videos generated from independent noise. However, this approach prohibits fine-grained comparisons, and we point out that it biases the annotators towards low-motion clips as they often contain fewer visual artifacts. In this work, we introduce DenseDPO, a method that addresses these shortcomings by making three contributions. First, we create each video pair for DPO by denoising corrupted copies of a ground truth video. This results in aligned pairs with similar motion structures while differing in local details, effectively neutralizing the motion bias. Second, we leverage the resulting temporal alignment to label preferences on short segments rather than entire clips, yielding a denser and more precise learning signal. With only one-third of the labeled data, DenseDPO greatly improves motion generation over vanilla DPO, while matching it in text alignment, visual quality, and temporal consistency. Finally, we show that DenseDPO unlocks automatic preference annotation using off-the-shelf Vision Language Models (VLMs): GPT accurately predicts segment-level preferences similar to task-specifically fine-tuned video reward models, and DenseDPO trained on these labels achieves performance close to using human labels. |
|
2025-06-05T00:00:00 | 2506.04142 | Establishing Trustworthy LLM Evaluation via Shortcut Neuron Analysis | [
"Kejian Zhu",
"Shangqing Tu",
"Zhuoran Jin",
"Lei Hou",
"Juanzi Li",
"Jun Zhao"
] | https://github.com/GaryStack/Trustworthy-Evaluation | The development of large language models (LLMs) depends on trustworthy evaluation. However, most current evaluations rely on public benchmarks, which are prone to data contamination issues that significantly compromise fairness. Previous researches have focused on constructing dynamic benchmarks to address contamination. However, continuously building new benchmarks is costly and cyclical. In this work, we aim to tackle contamination by analyzing the mechanisms of contaminated models themselves. Through our experiments, we discover that the overestimation of contaminated models is likely due to parameters acquiring shortcut solutions in training. We further propose a novel method for identifying shortcut neurons through comparative and causal analysis. Building on this, we introduce an evaluation method called shortcut neuron patching to suppress shortcut neurons. Experiments validate the effectiveness of our approach in mitigating contamination. Additionally, our evaluation results exhibit a strong linear correlation with MixEval, a recently released trustworthy benchmark, achieving a Spearman coefficient (rho) exceeding 0.95. This high correlation indicates that our method closely reveals true capabilities of the models and is trustworthy. We conduct further experiments to demonstrate the generalizability of our method across various benchmarks and hyperparameter settings. Code: https://github.com/GaryStack/Trustworthy-Evaluation |
2025-06-05T00:00:00 | 2506.04141 | MMR-V: What's Left Unsaid? A Benchmark for Multimodal Deep Reasoning in Videos | [
"Kejian Zhu",
"Zhuoran Jin",
"Hongbang Yuan",
"Jiachun Li",
"Shangqing Tu",
"Pengfei Cao",
"Yubo Chen",
"Kang Liu",
"Jun Zhao"
] | The sequential structure of videos poses a challenge to the ability of multimodal large language models (MLLMs) to locate multi-frame evidence and conduct multimodal reasoning. However, existing video benchmarks mainly focus on understanding tasks, which only require models to match frames mentioned in the question (hereafter referred to as "question frame") and perceive a few adjacent frames. To address this gap, we propose MMR-V: A Benchmark for Multimodal Deep Reasoning in Videos. The benchmark is characterized by the following features. (1) Long-range, multi-frame reasoning: Models are required to infer and analyze evidence frames that may be far from the question frame. (2) Beyond perception: Questions cannot be answered through direct perception alone but require reasoning over hidden information. (3) Reliability: All tasks are manually annotated, referencing extensive real-world user understanding to align with common perceptions. (4) Confusability: Carefully designed distractor annotation strategies to reduce model shortcuts. MMR-V consists of 317 videos and 1,257 tasks. Our experiments reveal that current models still struggle with multi-modal reasoning; even the best-performing model, o4-mini, achieves only 52.5% accuracy. Additionally, current reasoning enhancement strategies (Chain-of-Thought and scaling test-time compute) bring limited gains. Further analysis indicates that the CoT demanded for multi-modal reasoning differs from it in textual reasoning, which partly explains the limited performance gains. We hope that MMR-V can inspire further research into enhancing multi-modal reasoning capabilities. |
|
2025-06-05T00:00:00 | 2506.00482 | BenchHub: A Unified Benchmark Suite for Holistic and Customizable LLM Evaluation | [
"Eunsu Kim",
"Haneul Yoo",
"Guijin Son",
"Hitesh Patel",
"Amit Agarwal",
"Alice Oh"
] | As large language models (LLMs) continue to advance, the need for up-to-date and well-organized benchmarks becomes increasingly critical. However, many existing datasets are scattered, difficult to manage, and make it challenging to perform evaluations tailored to specific needs or domains, despite the growing importance of domain-specific models in areas such as math or code. In this paper, we introduce BenchHub, a dynamic benchmark repository that empowers researchers and developers to evaluate LLMs more effectively. BenchHub aggregates and automatically classifies benchmark datasets from diverse domains, integrating 303K questions across 38 benchmarks. It is designed to support continuous updates and scalable data management, enabling flexible and customizable evaluation tailored to various domains or use cases. Through extensive experiments with various LLM families, we demonstrate that model performance varies significantly across domain-specific subsets, emphasizing the importance of domain-aware benchmarking. We believe BenchHub can encourage better dataset reuse, more transparent model comparisons, and easier identification of underrepresented areas in existing benchmarks, offering a critical infrastructure for advancing LLM evaluation research. |
|
2025-06-05T00:00:00 | 2506.03956 | Adapt before Continual Learning | [
"Aojun Lu",
"Tao Feng",
"Hangjie Yuan",
"Chunhui Ding",
"Yanan Sun"
] | Continual Learning (CL) seeks to enable neural networks to incrementally acquire new knowledge (plasticity) while retaining existing knowledge (stability). While pre-trained models (PTMs) have become pivotal in CL, prevailing approaches freeze the PTM backbone to preserve stability, limiting their plasticity, particularly when encountering significant domain gaps in incremental tasks. Conversely, sequentially finetuning the entire PTM risks catastrophic forgetting of generalizable knowledge, exposing a critical stability-plasticity trade-off. To address this challenge, we propose Adapting PTMs before the core CL process (ACL), a novel framework that refines the PTM backbone through a plug-and-play adaptation phase before learning each new task with existing CL approaches (e.g., prompt tuning). ACL enhances plasticity by aligning embeddings with their original class prototypes while distancing them from others, theoretically and empirically shown to balance stability and plasticity. Extensive experiments demonstrate that ACL significantly improves CL performance across benchmarks and integrated methods, offering a versatile solution for PTM-based CL. |
|
2025-06-05T00:00:00 | 2506.03614 | VLMs Can Aggregate Scattered Training Patches | [
"Zhanhui Zhou",
"Lingjie Chen",
"Chao Yang",
"Chaochao Lu"
] | https://github.com/ZHZisZZ/visual-stitching | One way to mitigate risks in vision-language models (VLMs) is to remove dangerous samples in their training data. However, such data moderation can be easily bypassed when harmful images are split into small, benign-looking patches, scattered across many training samples. VLMs may then learn to piece these fragments together during training and generate harmful responses at inference, either from full images or text references. For instance, if trained on image patches from a bloody scene paired with the descriptions "safe," VLMs may later describe, the full image or a text reference to the scene, as "safe." We define the core ability of VLMs enabling this attack as visual stitching -- the ability to integrate visual information spread across multiple training samples that share the same textual descriptions. In our work, we first demonstrate visual stitching abilities in common open-source VLMs on three datasets where each image is labeled with a unique synthetic ID: we split each (image, ID) pair into {(patch, ID)} pairs at different granularity for finetuning, and we find that tuned models can verbalize the correct IDs from full images or text reference. Building on this, we simulate the adversarial data poisoning scenario mentioned above by using patches from dangerous images and replacing IDs with text descriptions like ``safe'' or ``unsafe'', demonstrating how harmful content can evade moderation in patches and later be reconstructed through visual stitching, posing serious VLM safety risks. Code is available at https://github.com/ZHZisZZ/visual-stitching. |
2025-06-05T00:00:00 | 2506.04225 | Voyager: Long-Range and World-Consistent Video Diffusion for Explorable 3D Scene Generation | [
"Tianyu Huang",
"Wangguandong Zheng",
"Tengfei Wang",
"Yuhao Liu",
"Zhenwei Wang",
"Junta Wu",
"Jie Jiang",
"Hui Li",
"Rynson W. H. Lau",
"Wangmeng Zuo",
"Chunchao Guo"
] | Real-world applications like video gaming and virtual reality often demand the ability to model 3D scenes that users can explore along custom camera trajectories. While significant progress has been made in generating 3D objects from text or images, creating long-range, 3D-consistent, explorable 3D scenes remains a complex and challenging problem. In this work, we present Voyager, a novel video diffusion framework that generates world-consistent 3D point-cloud sequences from a single image with user-defined camera path. Unlike existing approaches, Voyager achieves end-to-end scene generation and reconstruction with inherent consistency across frames, eliminating the need for 3D reconstruction pipelines (e.g., structure-from-motion or multi-view stereo). Our method integrates three key components: 1) World-Consistent Video Diffusion: A unified architecture that jointly generates aligned RGB and depth video sequences, conditioned on existing world observation to ensure global coherence 2) Long-Range World Exploration: An efficient world cache with point culling and an auto-regressive inference with smooth video sampling for iterative scene extension with context-aware consistency, and 3) Scalable Data Engine: A video reconstruction pipeline that automates camera pose estimation and metric depth prediction for arbitrary videos, enabling large-scale, diverse training data curation without manual 3D annotations. Collectively, these designs result in a clear improvement over existing methods in visual quality and geometric accuracy, with versatile applications. |
|
2025-06-05T00:00:00 | 2506.03930 | VisCoder: Fine-Tuning LLMs for Executable Python Visualization Code Generation | [
"Yuansheng Ni",
"Ping Nie",
"Kai Zou",
"Xiang Yue",
"Wenhu Chen"
] | Large language models (LLMs) often struggle with visualization tasks like plotting diagrams, charts, where success depends on both code correctness and visual semantics. Existing instruction-tuning datasets lack execution-grounded supervision and offer limited support for iterative code correction, resulting in fragile and unreliable plot generation. We present VisCode-200K, a large-scale instruction tuning dataset for Python-based visualization and self-correction. It contains over 200K examples from two sources: (1) validated plotting code from open-source repositories, paired with natural language instructions and rendered plots; and (2) 45K multi-turn correction dialogues from Code-Feedback, enabling models to revise faulty code using runtime feedback. We fine-tune Qwen2.5-Coder-Instruct on VisCode-200K to create VisCoder, and evaluate it on PandasPlotBench. VisCoder significantly outperforms strong open-source baselines and approaches the performance of proprietary models like GPT-4o-mini. We further adopt a self-debug evaluation protocol to assess iterative repair, demonstrating the benefits of feedback-driven learning for executable, visually accurate code generation. |
|
2025-06-05T00:00:00 | 2506.04089 | AmbiK: Dataset of Ambiguous Tasks in Kitchen Environment | [
"Anastasiia Ivanova",
"Eva Bakaeva",
"Zoya Volovikova",
"Alexey K. Kovalev",
"Aleksandr I. Panov"
] | https://github.com/cog-model/AmbiK-dataset | As a part of an embodied agent, Large Language Models (LLMs) are typically used for behavior planning given natural language instructions from the user. However, dealing with ambiguous instructions in real-world environments remains a challenge for LLMs. Various methods for task ambiguity detection have been proposed. However, it is difficult to compare them because they are tested on different datasets and there is no universal benchmark. For this reason, we propose AmbiK (Ambiguous Tasks in Kitchen Environment), the fully textual dataset of ambiguous instructions addressed to a robot in a kitchen environment. AmbiK was collected with the assistance of LLMs and is human-validated. It comprises 1000 pairs of ambiguous tasks and their unambiguous counterparts, categorized by ambiguity type (Human Preferences, Common Sense Knowledge, Safety), with environment descriptions, clarifying questions and answers, user intents, and task plans, for a total of 2000 tasks. We hope that AmbiK will enable researchers to perform a unified comparison of ambiguity detection methods. AmbiK is available at https://github.com/cog-model/AmbiK-dataset. |
2025-06-05T00:00:00 | 2506.03538 | Robust Neural Rendering in the Wild with Asymmetric Dual 3D Gaussian Splatting | [
"Chengqi Li",
"Zhihao Shi",
"Yangdi Lu",
"Wenbo He",
"Xiangyu Xu"
] | 3D reconstruction from in-the-wild images remains a challenging task due to inconsistent lighting conditions and transient distractors. Existing methods typically rely on heuristic strategies to handle the low-quality training data, which often struggle to produce stable and consistent reconstructions, frequently resulting in visual artifacts. In this work, we propose Asymmetric Dual 3DGS, a novel framework that leverages the stochastic nature of these artifacts: they tend to vary across different training runs due to minor randomness. Specifically, our method trains two 3D Gaussian Splatting (3DGS) models in parallel, enforcing a consistency constraint that encourages convergence on reliable scene geometry while suppressing inconsistent artifacts. To prevent the two models from collapsing into similar failure modes due to confirmation bias, we introduce a divergent masking strategy that applies two complementary masks: a multi-cue adaptive mask and a self-supervised soft mask, which leads to an asymmetric training process of the two models, reducing shared error modes. In addition, to improve the efficiency of model training, we introduce a lightweight variant called Dynamic EMA Proxy, which replaces one of the two models with a dynamically updated Exponential Moving Average (EMA) proxy, and employs an alternating masking strategy to preserve divergence. Extensive experiments on challenging real-world datasets demonstrate that our method consistently outperforms existing approaches while achieving high efficiency. Codes and trained models will be released. |
|
2025-06-05T00:00:00 | 2506.03951 | Rethinking the Stability-Plasticity Trade-off in Continual Learning from an Architectural Perspective | [
"Aojun Lu",
"Hangjie Yuan",
"Tao Feng",
"Yanan Sun"
] | The quest for Continual Learning (CL) seeks to empower neural networks with the ability to learn and adapt incrementally. Central to this pursuit is addressing the stability-plasticity dilemma, which involves striking a balance between two conflicting objectives: preserving previously learned knowledge and acquiring new knowledge. While numerous CL methods aim to achieve this trade-off, they often overlook the impact of network architecture on stability and plasticity, restricting the trade-off to the parameter level. In this paper, we delve into the conflict between stability and plasticity at the architectural level. We reveal that under an equal parameter constraint, deeper networks exhibit better plasticity, while wider networks are characterized by superior stability. To address this architectural-level dilemma, we introduce a novel framework denoted Dual-Arch, which serves as a plug-in component for CL. This framework leverages the complementary strengths of two distinct and independent networks: one dedicated to plasticity and the other to stability. Each network is designed with a specialized and lightweight architecture, tailored to its respective objective. Extensive experiments demonstrate that Dual-Arch enhances the performance of existing CL methods while being up to 87% more compact in terms of parameters. |
|
2025-06-05T00:00:00 | 2506.03355 | Robustness in Both Domains: CLIP Needs a Robust Text Encoder | [
"Elias Abad Rocamora",
"Christian Schlarmann",
"Naman Deep Singh",
"Yongtao Wu",
"Matthias Hein",
"Volkan Cevher"
] | Adversarial input attacks can cause a significant shift of CLIP embeddings. This can affect the downstream robustness of models incorporating CLIP in the pipeline, such as text-to-image generative models or large vision language models. While some efforts have been done towards making the CLIP image encoders robust, the robustness of text encoders remains unexplored. In this work, we cover this gap in the literature. We propose LEAF: an efficient adversarial finetuning method for the text domain, with the ability to scale to large CLIP models. Our models significantly improve the zero-shot adversarial accuracy in the text domain, while maintaining the vision performance provided by robust image encoders. When combined with text-to-image diffusion models, we can improve the generation quality under adversarial noise. When employing our robust CLIP encoders in multimodal retrieval tasks, we improve the recall under adversarial noise over standard CLIP models. Finally, we show that robust text encoders facilitate better reconstruction of input text from its embedding via direct optimization. |
|
2025-06-05T00:00:00 | 2506.03817 | Survey of Active Learning Hyperparameters: Insights from a Large-Scale Experimental Grid | [
"Julius Gonsior",
"Tim Rieß",
"Anja Reusch",
"Claudio Hartmann",
"Maik Thiele",
"Wolfgang Lehner"
] | Annotating data is a time-consuming and costly task, but it is inherently required for supervised machine learning. Active Learning (AL) is an established method that minimizes human labeling effort by iteratively selecting the most informative unlabeled samples for expert annotation, thereby improving the overall classification performance. Even though AL has been known for decades, AL is still rarely used in real-world applications. As indicated in the two community web surveys among the NLP community about AL, two main reasons continue to hold practitioners back from using AL: first, the complexity of setting AL up, and second, a lack of trust in its effectiveness. We hypothesize that both reasons share the same culprit: the large hyperparameter space of AL. This mostly unexplored hyperparameter space often leads to misleading and irreproducible AL experiment results. In this study, we first compiled a large hyperparameter grid of over 4.6 million hyperparameter combinations, second, recorded the performance of all combinations in the so-far biggest conducted AL study, and third, analyzed the impact of each hyperparameter in the experiment results. In the end, we give recommendations about the influence of each hyperparameter, demonstrate the surprising influence of the concrete AL strategy implementation, and outline an experimental study design for reproducible AL experiments with minimal computational effort, thus contributing to more reproducible and trustworthy AL research in the future. |
|
2025-06-05T00:00:00 | 2506.03150 | IllumiCraft: Unified Geometry and Illumination Diffusion for Controllable Video Generation | [
"Yuanze Lin",
"Yi-Wen Chen",
"Yi-Hsuan Tsai",
"Ronald Clark",
"Ming-Hsuan Yang"
] | Although diffusion-based models can generate high-quality and high-resolution video sequences from textual or image inputs, they lack explicit integration of geometric cues when controlling scene lighting and visual appearance across frames. To address this limitation, we propose IllumiCraft, an end-to-end diffusion framework accepting three complementary inputs: (1) high-dynamic-range (HDR) video maps for detailed lighting control; (2) synthetically relit frames with randomized illumination changes (optionally paired with a static background reference image) to provide appearance cues; and (3) 3D point tracks that capture precise 3D geometry information. By integrating the lighting, appearance, and geometry cues within a unified diffusion architecture, IllumiCraft generates temporally coherent videos aligned with user-defined prompts. It supports background-conditioned and text-conditioned video relighting and provides better fidelity than existing controllable video generation methods. Project Page: https://yuanze-lin.me/IllumiCraft_page |
|
2025-06-05T00:00:00 | 2506.02515 | FinChain: A Symbolic Benchmark for Verifiable Chain-of-Thought Financial Reasoning | [
"Zhuohan Xie",
"Dhruv Sahnan",
"Debopriyo Banerjee",
"Georgi Georgiev",
"Rushil Thareja",
"Hachem Madmoun",
"Jinyan Su",
"Aaryamonvikram Singh",
"Yuxia Wang",
"Rui Xing",
"Fajri Koto",
"Haonan Li",
"Ivan Koychev",
"Tanmoy Chakraborty",
"Salem Lahlou",
"Veselin Stoyanov",
"Preslav Nakov"
] | Multi-step symbolic reasoning is critical for advancing downstream performance on financial tasks. Yet, benchmarks for systematically evaluating this capability are lacking. Existing datasets like FinQA and ConvFinQA supervise only final numerical answers, without assessing intermediate reasoning steps. To address this, we introduce FinChain, the first symbolic benchmark designed for verifiable Chain-of- Thought (CoT) financial reasoning. Spanning 54 topics across 12 financial domains, Fin- Chain offers five parameterized templates per topic, each varying in reasoning complexity and domain expertise required. Each dataset instance includes an executable Python trace, enabling automatic generation of extensive training data and easy adaptation to other domains. We also introduce ChainEval, a new metric for automatic evaluation of both final answers and intermediate reasoning. Benchmarking 30 LLMs on our dataset, we find that even state-of-the-art models have considerable room for improvement in multi-step financial reasoning. All templates and evaluation metrics for FinChain are available at https: //github.com/mbzuai-nlp/finchain. |
|
2025-06-05T00:00:00 | 2506.02680 | Solving Inverse Problems with FLAIR | [
"Julius Erbach",
"Dominik Narnhofer",
"Andreas Dombos",
"Bernt Schiele",
"Jan Eric Lenssen",
"Konrad Schindler"
] | Flow-based latent generative models such as Stable Diffusion 3 are able to generate images with remarkable quality, even enabling photorealistic text-to-image generation. Their impressive performance suggests that these models should also constitute powerful priors for inverse imaging problems, but that approach has not yet led to comparable fidelity. There are several key obstacles: (i) the encoding into a lower-dimensional latent space makes the underlying (forward) mapping non-linear; (ii) the data likelihood term is usually intractable; and (iii) learned generative models struggle to recover rare, atypical data modes during inference. We present FLAIR, a novel training free variational framework that leverages flow-based generative models as a prior for inverse problems. To that end, we introduce a variational objective for flow matching that is agnostic to the type of degradation, and combine it with deterministic trajectory adjustments to recover atypical modes. To enforce exact consistency with the observed data, we decouple the optimization of the data fidelity and regularization terms. Moreover, we introduce a time-dependent calibration scheme in which the strength of the regularization is modulated according to off-line accuracy estimates. Results on standard imaging benchmarks demonstrate that FLAIR consistently outperforms existing diffusion- and flow-based methods in terms of reconstruction quality and sample diversity. |
|
2025-06-05T00:00:00 | 2506.03610 | Orak: A Foundational Benchmark for Training and Evaluating LLM Agents on Diverse Video Games | [
"Dongmin Park",
"Minkyu Kim",
"Beongjun Choi",
"Junhyuck Kim",
"Keon Lee",
"Jonghyun Lee",
"Inkyu Park",
"Byeong-Uk Lee",
"Jaeyoung Hwang",
"Jaewoo Ahn",
"Ameya S. Mahabaleshwarkar",
"Bilal Kartal",
"Pritam Biswas",
"Yoshi Suhara",
"Kangwook Lee",
"Jaewoong Cho"
] | https://github.com/krafton-ai/Orak | Large Language Model (LLM) agents are reshaping the game industry, particularly with more intelligent and human-preferable game characters. However, existing game benchmarks fall short of practical needs: they lack evaluations of diverse LLM capabilities across various game genres, studies of agentic modules crucial for complex gameplay, and fine-tuning datasets for aligning pre-trained LLMs into gaming agents. To fill these gaps, we present \benchname{}, a foundational benchmark designed to train and evaluate LLM agents across diverse real-world video games. Unlike existing benchmarks, Orak includes 12 popular video games spanning all major genres, enabling comprehensive studies of LLM capabilities and agentic modules essential for intricate game scenarios. To support consistent evaluation of LLMs, we introduce a plug-and-play interface based on Model Context Protocol (MCP) that enables LLMs to seamlessly connect with games and manipulate agentic modules. Additionally, we propose a fine-tuning dataset, consisting of LLM gameplay trajectories across diverse game genres. Orak offers a comprehensive evaluation framework, encompassing general game score leaderboards, LLM battle arenas, and in-depth analyses of visual input state, agentic strategies, and fine-tuning effects, establishing a foundation towards building generic gaming agents. Code is available at https://github.com/krafton-ai/Orak. |
2025-06-05T00:00:00 | 2506.03837 | HTSC-2025: A Benchmark Dataset of Ambient-Pressure High-Temperature Superconductors for AI-Driven Critical Temperature Prediction | [
"Xiao-Qi Han",
"Ze-Feng Gao",
"Xin-De Wang",
"Zhenfeng Ouyang",
"Peng-Jie Guo",
"Zhong-Yi Lu"
] | https://github.com/xqh19970407/HTSC-2025 | The discovery of high-temperature superconducting materials holds great significance for human industry and daily life. In recent years, research on predicting superconducting transition temperatures using artificial intelligence~(AI) has gained popularity, with most of these tools claiming to achieve remarkable accuracy. However, the lack of widely accepted benchmark datasets in this field has severely hindered fair comparisons between different AI algorithms and impeded further advancement of these methods. In this work, we present the HTSC-2025, an ambient-pressure high-temperature superconducting benchmark dataset. This comprehensive compilation encompasses theoretically predicted superconducting materials discovered by theoretical physicists from 2023 to 2025 based on BCS superconductivity theory, including the renowned X_2YH_6 system, perovskite MXH_3 system, M_3XH_8 system, cage-like BCN-doped metal atomic systems derived from LaH_{10} structural evolution, and two-dimensional honeycomb-structured systems evolving from MgB_2. The HTSC-2025 benchmark has been open-sourced at https://github.com/xqh19970407/HTSC-2025 and will be continuously updated. This benchmark holds significant importance for accelerating the discovery of superconducting materials using AI-based methods. |
2025-06-05T00:00:00 | 2506.03525 | Video-Skill-CoT: Skill-based Chain-of-Thoughts for Domain-Adaptive Video Reasoning | [
"Daeun Lee",
"Jaehong Yoon",
"Jaemin Cho",
"Mohit Bansal"
] | Recent advances in Chain-of-Thought (CoT) reasoning have improved complex video understanding, but existing methods often struggle to adapt to domain-specific skills (e.g., event detection, spatial relation understanding, emotion understanding) over various video content. To address this, we propose Video-Skill-CoT (a.k.a. Video-SKoT), a framework that automatically constructs and leverages skill-aware CoT supervisions for domain-adaptive video reasoning. First, we construct skill-based CoT annotations: we extract domain-relevant reasoning skills from training questions, cluster them into a shared skill taxonomy, and create detailed multi-step CoT rationale tailored to each video-question pair for training. Second, we introduce a skill-specific expert learning framework. Each expert module specializes in a subset of reasoning skills and is trained with lightweight adapters using the collected CoT supervision. We demonstrate the effectiveness of the proposed approach on three video understanding benchmarks, where Video-SKoT consistently outperforms strong baselines. We also provide in-depth analyses on comparing different CoT annotation pipelines and learned skills over multiple video domains. |
|
2025-06-05T00:00:00 | 2505.23564 | Segment Policy Optimization: Effective Segment-Level Credit Assignment in RL for Large Language Models | [
"Yiran Guo",
"Lijie Xu",
"Jie Liu",
"Dan Ye",
"Shuang Qiu"
] | https://github.com/AIFrameResearch/SPO | Enhancing the reasoning capabilities of large language models effectively using reinforcement learning (RL) remains a crucial challenge. Existing approaches primarily adopt two contrasting advantage estimation granularities: Token-level methods (e.g., PPO) aim to provide the fine-grained advantage signals but suffer from inaccurate estimation due to difficulties in training an accurate critic model. On the other extreme, trajectory-level methods (e.g., GRPO) solely rely on a coarse-grained advantage signal from the final reward, leading to imprecise credit assignment. To address these limitations, we propose Segment Policy Optimization (SPO), a novel RL framework that leverages segment-level advantage estimation at an intermediate granularity, achieving a better balance by offering more precise credit assignment than trajectory-level methods and requiring fewer estimation points than token-level methods, enabling accurate advantage estimation based on Monte Carlo (MC) without a critic model. SPO features three components with novel strategies: (1) flexible segment partition; (2) accurate segment advantage estimation; and (3) policy optimization using segment advantages, including a novel probability-mask strategy. We further instantiate SPO for two specific scenarios: (1) SPO-chain for short chain-of-thought (CoT), featuring novel cutpoint-based partition and chain-based advantage estimation, achieving 6-12 percentage point improvements in accuracy over PPO and GRPO on GSM8K. (2) SPO-tree for long CoT, featuring novel tree-based advantage estimation, which significantly reduces the cost of MC estimation, achieving 7-11 percentage point improvements over GRPO on MATH500 under 2K and 4K context evaluation. We make our code publicly available at https://github.com/AIFrameResearch/SPO. |
2025-06-05T00:00:00 | 2506.04214 | Sounding that Object: Interactive Object-Aware Image to Audio Generation | [
"Tingle Li",
"Baihe Huang",
"Xiaobin Zhuang",
"Dongya Jia",
"Jiawei Chen",
"Yuping Wang",
"Zhuo Chen",
"Gopala Anumanchipalli",
"Yuxuan Wang"
] | Generating accurate sounds for complex audio-visual scenes is challenging, especially in the presence of multiple objects and sound sources. In this paper, we propose an {\em interactive object-aware audio generation} model that grounds sound generation in user-selected visual objects within images. Our method integrates object-centric learning into a conditional latent diffusion model, which learns to associate image regions with their corresponding sounds through multi-modal attention. At test time, our model employs image segmentation to allow users to interactively generate sounds at the {\em object} level. We theoretically validate that our attention mechanism functionally approximates test-time segmentation masks, ensuring the generated audio aligns with selected objects. Quantitative and qualitative evaluations show that our model outperforms baselines, achieving better alignment between objects and their associated sounds. Project page: https://tinglok.netlify.app/files/avobject/ |
|
2025-06-05T00:00:00 | 2506.03822 | CRAWLDoc: A Dataset for Robust Ranking of Bibliographic Documents | [
"Fabian Karl",
"Ansgar Scherp"
] | https://github.com/FKarl/CRAWLDoc | Publication databases rely on accurate metadata extraction from diverse web sources, yet variations in web layouts and data formats present challenges for metadata providers. This paper introduces CRAWLDoc, a new method for contextual ranking of linked web documents. Starting with a publication's URL, such as a digital object identifier, CRAWLDoc retrieves the landing page and all linked web resources, including PDFs, ORCID profiles, and supplementary materials. It embeds these resources, along with anchor texts and the URLs, into a unified representation. For evaluating CRAWLDoc, we have created a new, manually labeled dataset of 600 publications from six top publishers in computer science. Our method CRAWLDoc demonstrates a robust and layout-independent ranking of relevant documents across publishers and data formats. It lays the foundation for improved metadata extraction from web documents with various layouts and formats. Our source code and dataset can be accessed at https://github.com/FKarl/CRAWLDoc. |
2025-06-05T00:00:00 | 2506.02153 | Small Language Models are the Future of Agentic AI | [
"Peter Belcak",
"Greg Heinrich",
"Shizhe Diao",
"Yonggan Fu",
"Xin Dong",
"Saurav Muralidharan",
"Yingyan Celine Lin",
"Pavlo Molchanov"
] | Large language models (LLMs) are often praised for exhibiting near-human performance on a wide range of tasks and valued for their ability to hold a general conversation. The rise of agentic AI systems is, however, ushering in a mass of applications in which language models perform a small number of specialized tasks repetitively and with little variation. Here we lay out the position that small language models (SLMs) are sufficiently powerful, inherently more suitable, and necessarily more economical for many invocations in agentic systems, and are therefore the future of agentic AI. Our argumentation is grounded in the current level of capabilities exhibited by SLMs, the common architectures of agentic systems, and the economy of LM deployment. We further argue that in situations where general-purpose conversational abilities are essential, heterogeneous agentic systems (i.e., agents invoking multiple different models) are the natural choice. We discuss the potential barriers for the adoption of SLMs in agentic systems and outline a general LLM-to-SLM agent conversion algorithm. Our position, formulated as a value statement, highlights the significance of the operational and economic impact even a partial shift from LLMs to SLMs is to have on the AI agent industry. We aim to stimulate the discussion on the effective use of AI resources and hope to advance the efforts to lower the costs of AI of the present day. Calling for both contributions to and critique of our position, we commit to publishing all such correspondence at https://research.nvidia.com/labs/lpr/slm-agents. |
|
2025-06-05T00:00:00 | 2506.00618 | RiOSWorld: Benchmarking the Risk of Multimodal Compter-Use Agents | [
"Jingyi Yang",
"Shuai Shao",
"Dongrui Liu",
"Jing Shao"
] | With the rapid development of multimodal large language models (MLLMs), they are increasingly deployed as autonomous computer-use agents capable of accomplishing complex computer tasks. However, a pressing issue arises: Can the safety risk principles designed and aligned for general MLLMs in dialogue scenarios be effectively transferred to real-world computer-use scenarios? Existing research on evaluating the safety risks of MLLM-based computer-use agents suffers from several limitations: it either lacks realistic interactive environments, or narrowly focuses on one or a few specific risk types. These limitations ignore the complexity, variability, and diversity of real-world environments, thereby restricting comprehensive risk evaluation for computer-use agents. To this end, we introduce RiOSWorld, a benchmark designed to evaluate the potential risks of MLLM-based agents during real-world computer manipulations. Our benchmark includes 492 risky tasks spanning various computer applications, involving web, social media, multimedia, os, email, and office software. We categorize these risks into two major classes based on their risk source: (i) User-originated risks and (ii) Environmental risks. For the evaluation, we evaluate safety risks from two perspectives: (i) Risk goal intention and (ii) Risk goal completion. Extensive experiments with multimodal agents on RiOSWorld demonstrate that current computer-use agents confront significant safety risks in real-world scenarios. Our findings highlight the necessity and urgency of safety alignment for computer-use agents in real-world computer manipulation, providing valuable insights for developing trustworthy computer-use agents. Our benchmark is publicly available at https://yjyddq.github.io/RiOSWorld.github.io/. |
|
2025-06-05T00:00:00 | 2506.04178 | OpenThoughts: Data Recipes for Reasoning Models | [
"Etash Guha",
"Ryan Marten",
"Sedrick Keh",
"Negin Raoof",
"Georgios Smyrnis",
"Hritik Bansal",
"Marianna Nezhurina",
"Jean Mercat",
"Trung Vu",
"Zayne Sprague",
"Ashima Suvarna",
"Benjamin Feuer",
"Liangyu Chen",
"Zaid Khan",
"Eric Frankel",
"Sachin Grover",
"Caroline Choi",
"Niklas Muennighoff",
"Shiye Su",
"Wanjia Zhao",
"John Yang",
"Shreyas Pimpalgaonkar",
"Kartik Sharma",
"Charlie Cheng-Jie Ji",
"Yichuan Deng",
"Sarah Pratt",
"Vivek Ramanujan",
"Jon Saad-Falcon",
"Jeffrey Li",
"Achal Dave",
"Alon Albalak",
"Kushal Arora",
"Blake Wulfe",
"Chinmay Hegde",
"Greg Durrett",
"Sewoong Oh",
"Mohit Bansal",
"Saadia Gabriel",
"Aditya Grover",
"Kai-Wei Chang",
"Vaishaal Shankar",
"Aaron Gokaslan",
"Mike A. Merrill",
"Tatsunori Hashimoto",
"Yejin Choi",
"Jenia Jitsev",
"Reinhard Heckel",
"Maheswaran Sathiamoorthy",
"Alexandros G. Dimakis",
"Ludwig Schmidt"
] | Reasoning models have made rapid progress on many benchmarks involving math, code, and science. Yet, there are still many open questions about the best training recipes for reasoning since state-of-the-art models often rely on proprietary datasets with little to no public information available. To address this, the goal of the OpenThoughts project is to create open-source datasets for training reasoning models. After initial explorations, our OpenThoughts2-1M dataset led to OpenThinker2-32B, the first model trained on public reasoning data to match DeepSeek-R1-Distill-32B on standard reasoning benchmarks such as AIME and LiveCodeBench. We then improve our dataset further by systematically investigating each step of our data generation pipeline with 1,000+ controlled experiments, which led to OpenThoughts3. Scaling the pipeline to 1.2M examples and using QwQ-32B as teacher yields our OpenThinker3-7B model, which achieves state-of-the-art results: 53% on AIME 2025, 51% on LiveCodeBench 06/24-01/25, and 54% on GPQA Diamond. All of our datasets and models are available on https://openthoughts.ai. |
|
2025-06-05T00:00:00 | 2506.03566 | POSS: Position Specialist Generates Better Draft for Speculative Decoding | [
"Langlin Huang",
"Chengsong Huang",
"Jixuan Leng",
"Di Huang",
"Jiaxin Huang"
] | https://github.com/shrango/PosS | Speculative decoding accelerates Large Language Model (LLM) inference by using a small draft model to predict multiple tokens, and a large target model to verify these tokens in parallel. Recent studies leverage the hidden state of the target model to enhance draft model prediction accuracy. However, existing methods suffer from the degrading quality of draft token predictions at later positions, due to error accumulation in draft model generated features. In this paper, we propose Position Specialists (PosS), which consist of multiple position-specialized draft layers to generate tokens at assigned position(s). Position specialists greatly improve token acceptance rate at later positions per drafting round, as each specialist only needs to focus on handling a certain level of draft model feature deviation. Experiment results on Llama-3-8B-Instruct and Llama-2-13B-chat across six datasets demonstrate that PosS effectively improves over baselines on average acceptance length and speed-up ratio. Our codebase is available at https://github.com/shrango/PosS. |
2025-06-05T00:00:00 | 2506.02863 | CapSpeech: Enabling Downstream Applications in Style-Captioned Text-to-Speech | [
"Helin Wang",
"Jiarui Hai",
"Dading Chong",
"Karan Thakkar",
"Tiantian Feng",
"Dongchao Yang",
"Junhyeok Lee",
"Laureano Moro Velazquez",
"Jesus Villalba",
"Zengyi Qin",
"Shrikanth Narayanan",
"Mounya Elhiali",
"Najim Dehak"
] | Recent advancements in generative artificial intelligence have significantly transformed the field of style-captioned text-to-speech synthesis (CapTTS). However, adapting CapTTS to real-world applications remains challenging due to the lack of standardized, comprehensive datasets and limited research on downstream tasks built upon CapTTS. To address these gaps, we introduce CapSpeech, a new benchmark designed for a series of CapTTS-related tasks, including style-captioned text-to-speech synthesis with sound events (CapTTS-SE), accent-captioned TTS (AccCapTTS), emotion-captioned TTS (EmoCapTTS), and text-to-speech synthesis for chat agent (AgentTTS). CapSpeech comprises over 10 million machine-annotated audio-caption pairs and nearly 0.36 million human-annotated audio-caption pairs. In addition, we introduce two new datasets collected and recorded by a professional voice actor and experienced audio engineers, specifically for the AgentTTS and CapTTS-SE tasks. Alongside the datasets, we conduct comprehensive experiments using both autoregressive and non-autoregressive models on CapSpeech. Our results demonstrate high-fidelity and highly intelligible speech synthesis across a diverse range of speaking styles. To the best of our knowledge, CapSpeech is the largest available dataset offering comprehensive annotations for CapTTS-related tasks. The experiments and findings further provide valuable insights into the challenges of developing CapTTS systems. |
|
2025-06-05T00:00:00 | 2506.05332 | Unleashing Hour-Scale Video Training for Long Video-Language Understanding | [
"Jingyang Lin",
"Jialian Wu",
"Ximeng Sun",
"Ze Wang",
"Jiang Liu",
"Yusheng Su",
"Xiaodong Yu",
"Hao Chen",
"Jiebo Luo",
"Zicheng Liu",
"Emad Barsoum"
] | Recent long-form video-language understanding benchmarks have driven progress in video large multimodal models (Video-LMMs). However, the scarcity of well-annotated long videos has left the training of hour-long Video-LLMs underexplored. To close this gap, we present VideoMarathon, a large-scale hour-long video instruction-following dataset. This dataset includes around 9,700 hours of long videos sourced from diverse domains, ranging from 3 to 60 minutes per video. Specifically, it contains 3.3M high-quality QA pairs, spanning six fundamental topics: temporality, spatiality, object, action, scene, and event. Compared to existing video instruction datasets, VideoMarathon significantly extends training video durations up to 1 hour, and supports 22 diverse tasks requiring both short- and long-term video comprehension. Building on VideoMarathon, we propose Hour-LLaVA, a powerful and efficient Video-LMM for hour-scale video-language modeling. It enables hour-long video training and inference at 1-FPS sampling by leveraging a memory augmentation module, which adaptively integrates user question-relevant and spatiotemporal-informative semantics from a cached full video context. In our experiments, Hour-LLaVA achieves the best performance on multiple long video-language benchmarks, demonstrating the high quality of the VideoMarathon dataset and the superiority of the Hour-LLaVA model. |
|
2025-06-06T00:00:00 | 2506.05328 | AV-Reasoner: Improving and Benchmarking Clue-Grounded Audio-Visual Counting for MLLMs | [
"Lidong Lu",
"Guo Chen",
"Zhiqi Li",
"Yicheng Liu",
"Tong Lu"
] | Despite progress in video understanding, current MLLMs struggle with counting tasks. Existing benchmarks are limited by short videos, close-set queries, lack of clue annotations, and weak multimodal coverage. In this paper, we introduce CG-AV-Counting, a manually-annotated clue-grounded counting benchmark with 1,027 multimodal questions and 5,845 annotated clues over 497 long videos. It supports both black-box and white-box evaluation, serving as a comprehensive testbed for both end-to-end and reasoning-based counting. To explore ways to improve model's counting capability, we propose AV-Reasoner, a model trained with GRPO and curriculum learning to generalize counting ability from related tasks. AV-Reasoner achieves state-of-the-art results across multiple benchmarks, demonstrating the effectiveness of reinforcement learning. However, experiments show that on out-of-domain benchmarks, reasoning in the language space fails to bring performance gains. The code and benchmark have been realeased on https://av-reasoner.github.io. |
|
2025-06-06T00:00:00 | 2506.04308 | RoboRefer: Towards Spatial Referring with Reasoning in Vision-Language Models for Robotics | [
"Enshen Zhou",
"Jingkun An",
"Cheng Chi",
"Yi Han",
"Shanyu Rong",
"Chi Zhang",
"Pengwei Wang",
"Zhongyuan Wang",
"Tiejun Huang",
"Lu Sheng",
"Shanghang Zhang"
] | Spatial referring is a fundamental capability of embodied robots to interact with the 3D physical world. However, even with the powerful pretrained vision language models (VLMs), recent approaches are still not qualified to accurately understand the complex 3D scenes and dynamically reason about the instruction-indicated locations for interaction. To this end, we propose RoboRefer, a 3D-aware VLM that can first achieve precise spatial understanding by integrating a disentangled but dedicated depth encoder via supervised fine-tuning (SFT). Moreover, RoboRefer advances generalized multi-step spatial reasoning via reinforcement fine-tuning (RFT), with metric-sensitive process reward functions tailored for spatial referring tasks. To support SFT and RFT training, we introduce RefSpatial, a large-scale dataset of 20M QA pairs (2x prior), covering 31 spatial relations (vs. 15 prior) and supporting complex reasoning processes (up to 5 steps). In addition, we introduce RefSpatial-Bench, a challenging benchmark filling the gap in evaluating spatial referring with multi-step reasoning. Experiments show that SFT-trained RoboRefer achieves state-of-the-art spatial understanding, with an average success rate of 89.6%. RFT-trained RoboRefer further outperforms all other baselines by a large margin, even surpassing Gemini-2.5-Pro by 17.4% in average accuracy on RefSpatial-Bench. Notably, RoboRefer can be integrated with various control policies to execute long-horizon, dynamic tasks across diverse robots (e,g., UR5, G1 humanoid) in cluttered real-world scenes. |
|
2025-06-06T00:00:00 | 2505.23656 | VideoREPA: Learning Physics for Video Generation through Relational Alignment with Foundation Models | [
"Xiangdong Zhang",
"Jiaqi Liao",
"Shaofeng Zhang",
"Fanqing Meng",
"Xiangpeng Wan",
"Junchi Yan",
"Yu Cheng"
] | Recent advancements in text-to-video (T2V) diffusion models have enabled high-fidelity and realistic video synthesis. However, current T2V models often struggle to generate physically plausible content due to their limited inherent ability to accurately understand physics. We found that while the representations within T2V models possess some capacity for physics understanding, they lag significantly behind those from recent video self-supervised learning methods. To this end, we propose a novel framework called VideoREPA, which distills physics understanding capability from video understanding foundation models into T2V models by aligning token-level relations. This closes the physics understanding gap and enable more physics-plausible generation. Specifically, we introduce the Token Relation Distillation (TRD) loss, leveraging spatio-temporal alignment to provide soft guidance suitable for finetuning powerful pre-trained T2V models, a critical departure from prior representation alignment (REPA) methods. To our knowledge, VideoREPA is the first REPA method designed for finetuning T2V models and specifically for injecting physical knowledge. Empirical evaluations show that VideoREPA substantially enhances the physics commonsense of baseline method, CogVideoX, achieving significant improvement on relevant benchmarks and demonstrating a strong capacity for generating videos consistent with intuitive physics. More video results are available at https://videorepa.github.io/. |
|
2025-06-06T00:00:00 | 2506.04209 | Language-Image Alignment with Fixed Text Encoders | [
"Jingfeng Yang",
"Ziyang Wu",
"Yue Zhao",
"Yi Ma"
] | Currently, the most dominant approach to establishing language-image alignment is to pre-train text and image encoders jointly through contrastive learning, such as CLIP and its variants. In this work, we question whether such a costly joint training is necessary. In particular, we investigate if a pre-trained fixed large language model (LLM) offers a good enough text encoder to guide visual representation learning. That is, we propose to learn Language-Image alignment with a Fixed Text encoder (LIFT) from an LLM by training only the image encoder. Somewhat surprisingly, through comprehensive benchmarking and ablation studies, we find that this much simplified framework LIFT is highly effective and it outperforms CLIP in most scenarios that involve compositional understanding and long captions, while achieving considerable gains in computational efficiency. Our work takes a first step towards systematically exploring how text embeddings from LLMs can guide visual learning and suggests an alternative design choice for learning language-aligned visual representations. |
|
2025-06-06T00:00:00 | 2506.05176 | Qwen3 Embedding: Advancing Text Embedding and Reranking Through Foundation Models | [
"Yanzhao Zhang",
"Mingxin Li",
"Dingkun Long",
"Xin Zhang",
"Huan Lin",
"Baosong Yang",
"Pengjun Xie",
"An Yang",
"Dayiheng Liu",
"Junyang Lin",
"Fei Huang",
"Jingren Zhou"
] | In this work, we introduce the Qwen3 Embedding series, a significant advancement over its predecessor, the GTE-Qwen series, in text embedding and reranking capabilities, built upon the Qwen3 foundation models. Leveraging the Qwen3 LLMs' robust capabilities in multilingual text understanding and generation, our innovative multi-stage training pipeline combines large-scale unsupervised pre-training with supervised fine-tuning on high-quality datasets. Effective model merging strategies further ensure the robustness and adaptability of the Qwen3 Embedding series. During the training process, the Qwen3 LLMs serve not only as backbone models but also play a crucial role in synthesizing high-quality, rich, and diverse training data across multiple domains and languages, thus enhancing the training pipeline. The Qwen3 Embedding series offers a spectrum of model sizes (0.6B, 4B, 8B) for both embedding and reranking tasks, addressing diverse deployment scenarios where users can optimize for either efficiency or effectiveness. Empirical evaluations demonstrate that the Qwen3 Embedding series achieves state-of-the-art results across diverse benchmarks. Notably, it excels on the multilingual evaluation benchmark MTEB for text embedding, as well as in various retrieval tasks, including code retrieval, cross-lingual retrieval and multilingual retrieval. To facilitate reproducibility and promote community-driven research and development, the Qwen3 Embedding models are publicly available under the Apache 2.0 license. |
|
2025-06-06T00:00:00 | 2506.05344 | SparseMM: Head Sparsity Emerges from Visual Concept Responses in MLLMs | [
"Jiahui Wang",
"Zuyan Liu",
"Yongming Rao",
"Jiwen Lu"
] | https://github.com/CR400AF-A/SparseMM | Multimodal Large Language Models (MLLMs) are commonly derived by extending pre-trained Large Language Models (LLMs) with visual capabilities. In this work, we investigate how MLLMs process visual inputs by analyzing their attention mechanisms. We reveal a surprising sparsity phenomenon: only a small subset (approximately less than 5%) of attention heads in LLMs actively contribute to visual understanding, termed visual heads. To identify these heads efficiently, we design a training-free framework that quantifies head-level visual relevance through targeted response analysis. Building on this discovery, we introduce SparseMM, a KV-Cache optimization strategy that allocates asymmetric computation budgets to heads in LLMs based on their visual scores, leveraging the sparity of visual heads for accelerating the inference of MLLMs. Compared with prior KV-Cache acceleration methods that ignore the particularity of visual, SparseMM prioritizes stress and retaining visual semantics during decoding. Extensive evaluations across mainstream multimodal benchmarks demonstrate that SparseMM achieves superior accuracy-efficiency trade-offs. Notably, SparseMM delivers 1.38x real-time acceleration and 52% memory reduction during generation while maintaining performance parity on efficiency test. Our project is open sourced at https://github.com/CR400AF-A/SparseMM. |
2025-06-06T00:00:00 | 2506.05240 | Aligning Latent Spaces with Flow Priors | [
"Yizhuo Li",
"Yuying Ge",
"Yixiao Ge",
"Ying Shan",
"Ping Luo"
] | This paper presents a novel framework for aligning learnable latent spaces to arbitrary target distributions by leveraging flow-based generative models as priors. Our method first pretrains a flow model on the target features to capture the underlying distribution. This fixed flow model subsequently regularizes the latent space via an alignment loss, which reformulates the flow matching objective to treat the latents as optimization targets. We formally prove that minimizing this alignment loss establishes a computationally tractable surrogate objective for maximizing a variational lower bound on the log-likelihood of latents under the target distribution. Notably, the proposed method eliminates computationally expensive likelihood evaluations and avoids ODE solving during optimization. As a proof of concept, we demonstrate in a controlled setting that the alignment loss landscape closely approximates the negative log-likelihood of the target distribution. We further validate the effectiveness of our approach through large-scale image generation experiments on ImageNet with diverse target distributions, accompanied by detailed discussions and ablation studies. With both theoretical and empirical validation, our framework paves a new way for latent space alignment. |
|
2025-06-06T00:00:00 | 2506.04405 | MedAgentGym: Training LLM Agents for Code-Based Medical Reasoning at Scale | [
"Ran Xu",
"Yuchen Zhuang",
"Yishan Zhong",
"Yue Yu",
"Xiangru Tang",
"Hang Wu",
"May D. Wang",
"Peifeng Ruan",
"Donghan Yang",
"Tao Wang",
"Guanghua Xiao",
"Carl Yang",
"Yang Xie",
"Wenqi Shi"
] | We introduce MedAgentGYM, the first publicly available training environment designed to enhance coding-based medical reasoning capabilities in large language model (LLM) agents. MedAgentGYM comprises 72,413 task instances across 129 categories derived from authentic real-world biomedical scenarios. Tasks are encapsulated within executable coding environments, each featuring detailed task descriptions, interactive feedback mechanisms, verifiable ground-truth annotations, and scalable training trajectory generation. Extensive benchmarking of over 30 LLMs reveals a notable performance disparity between commercial API-based models and open-source counterparts. Leveraging MedAgentGYM, Med-Copilot-7B achieves substantial performance gains through supervised fine-tuning (+36.44%) and continued reinforcement learning (+42.47%), emerging as an affordable and privacy-preserving alternative competitive with gpt-4o. By offering both a comprehensive benchmark and accessible, expandable training resources within unified execution environments, MedAgentGYM delivers an integrated platform to develop LLM-based coding assistants for advanced biomedical research and practice. |
|
2025-06-06T00:00:00 | 2506.04245 | Contextual Integrity in LLMs via Reasoning and Reinforcement Learning | [
"Guangchen Lan",
"Huseyin A. Inan",
"Sahar Abdelnabi",
"Janardhan Kulkarni",
"Lukas Wutschitz",
"Reza Shokri",
"Christopher G. Brinton",
"Robert Sim"
] | As the era of autonomous agents making decisions on behalf of users unfolds, ensuring contextual integrity (CI) -- what is the appropriate information to share while carrying out a certain task -- becomes a central question to the field. We posit that CI demands a form of reasoning where the agent needs to reason about the context in which it is operating. To test this, we first prompt LLMs to reason explicitly about CI when deciding what information to disclose. We then extend this approach by developing a reinforcement learning (RL) framework that further instills in models the reasoning necessary to achieve CI. Using a synthetic, automatically created, dataset of only sim700 examples but with diverse contexts and information disclosure norms, we show that our method substantially reduces inappropriate information disclosure while maintaining task performance across multiple model sizes and families. Importantly, improvements transfer from this synthetic dataset to established CI benchmarks such as PrivacyLens that has human annotations and evaluates privacy leakage of AI assistants in actions and tool calls. |
|
2025-06-06T00:00:00 | 2506.04734 | Evaluation is All You Need: Strategic Overclaiming of LLM Reasoning Capabilities Through Evaluation Design | [
"Lin Sun",
"Weihong Lin",
"Jinzhu Wu",
"Yongfu Zhu",
"Xiaoqi Jian",
"Guangxiang Zhao",
"Change Jia",
"Linglin Zhang",
"Sai-er Hu",
"Yuhan Wu",
"Xiangzheng Zhang"
] | Reasoning models represented by the Deepseek-R1-Distill series have been widely adopted by the open-source community due to their strong performance in mathematics, science, programming, and other domains. However, our study reveals that their benchmark evaluation results are subject to significant fluctuations caused by various factors. Subtle differences in evaluation conditions can lead to substantial variations in results. Similar phenomena are observed in other open-source inference models fine-tuned based on the Deepseek-R1-Distill series, as well as in the QwQ-32B model, making their claimed performance improvements difficult to reproduce reliably. Therefore, we advocate for the establishment of a more rigorous paradigm for model performance evaluation and present our empirical assessments of the Deepseek-R1-Distill series models. |
|
2025-06-06T00:00:00 | 2506.03238 | Rethinking Whole-Body CT Image Interpretation: An Abnormality-Centric Approach | [
"Ziheng Zhao",
"Lisong Dai",
"Ya Zhang",
"Yanfeng Wang",
"Weidi Xie"
] | Automated interpretation of CT images-particularly localizing and describing abnormal findings across multi-plane and whole-body scans-remains a significant challenge in clinical radiology. This work aims to address this challenge through four key contributions: (i) On taxonomy, we collaborate with senior radiologists to propose a comprehensive hierarchical classification system, with 404 representative abnormal findings across all body regions; (ii) On data, we contribute a dataset containing over 14.5K CT images from multiple planes and all human body regions, and meticulously provide grounding annotations for over 19K abnormalities, each linked to the detailed description and cast into the taxonomy; (iii) On model development, we propose OminiAbnorm-CT, which can automatically ground and describe abnormal findings on multi-plane and whole-body CT images based on text queries, while also allowing flexible interaction through visual prompts; (iv) On benchmarks, we establish three representative evaluation tasks based on real clinical scenarios. Through extensive experiments, we show that OminiAbnorm-CT can significantly outperform existing methods on all the tasks and metrics. |
|
2025-06-06T00:00:00 | 2506.05287 | EOC-Bench: Can MLLMs Identify, Recall, and Forecast Objects in an Egocentric World? | [
"Yuqian Yuan",
"Ronghao Dang",
"Long Li",
"Wentong Li",
"Dian Jiao",
"Xin Li",
"Deli Zhao",
"Fan Wang",
"Wenqiao Zhang",
"Jun Xiao",
"Yueting Zhuang"
] | The emergence of multimodal large language models (MLLMs) has driven breakthroughs in egocentric vision applications. These applications necessitate persistent, context-aware understanding of objects, as users interact with tools in dynamic and cluttered environments. However, existing embodied benchmarks primarily focus on static scene exploration, emphasizing object's appearance and spatial attributes while neglecting the assessment of dynamic changes arising from users' interactions. To address this gap, we introduce EOC-Bench, an innovative benchmark designed to systematically evaluate object-centric embodied cognition in dynamic egocentric scenarios. Specially, EOC-Bench features 3,277 meticulously annotated QA pairs categorized into three temporal categories: Past, Present, and Future, covering 11 fine-grained evaluation dimensions and 3 visual object referencing types. To ensure thorough assessment, we develop a mixed-format human-in-the-loop annotation framework with four types of questions and design a novel multi-scale temporal accuracy metric for open-ended temporal evaluation. Based on EOC-Bench, we conduct comprehensive evaluations of various proprietary, open-source, and object-level MLLMs. EOC-Bench serves as a crucial tool for advancing the embodied object cognitive capabilities of MLLMs, establishing a robust foundation for developing reliable core models for embodied systems. |
|
2025-06-06T00:00:00 | 2506.05327 | Revisiting Depth Representations for Feed-Forward 3D Gaussian Splatting | [
"Duochao Shi",
"Weijie Wang",
"Donny Y. Chen",
"Zeyu Zhang",
"Jia-Wang Bian",
"Bohan Zhuang",
"Chunhua Shen"
] | Depth maps are widely used in feed-forward 3D Gaussian Splatting (3DGS) pipelines by unprojecting them into 3D point clouds for novel view synthesis. This approach offers advantages such as efficient training, the use of known camera poses, and accurate geometry estimation. However, depth discontinuities at object boundaries often lead to fragmented or sparse point clouds, degrading rendering quality -- a well-known limitation of depth-based representations. To tackle this issue, we introduce PM-Loss, a novel regularization loss based on a pointmap predicted by a pre-trained transformer. Although the pointmap itself may be less accurate than the depth map, it effectively enforces geometric smoothness, especially around object boundaries. With the improved depth map, our method significantly improves the feed-forward 3DGS across various architectures and scenes, delivering consistently better rendering results. Our project page: https://aim-uofa.github.io/PMLoss |
|
2025-06-06T00:00:00 | 2506.03077 | StreamBP: Memory-Efficient Exact Backpropagation for Long Sequence Training of LLMs | [
"Qijun Luo",
"Mengqi Li",
"Lei Zhao",
"Xiao Li"
] | https://github.com/Ledzy/StreamBP | Training language models on long sequence data is a demanding requirement for enhancing the model's capability on complex tasks, e.g., long-chain reasoning. However, as the sequence length scales up, the memory cost for storing activation values becomes huge during the Backpropagation (BP) process, even with the application of gradient checkpointing technique. To tackle this challenge, we propose a memory-efficient and exact BP method called StreamBP, which performs a linear decomposition of the chain rule along the sequence dimension in a layer-wise manner, significantly reducing the memory cost of activation values and logits. The proposed method is applicable to common objectives such as SFT, GRPO, and DPO. From an implementation perspective, StreamBP achieves less computational FLOPs and faster BP speed by leveraging the causal structure of the language model. Compared to gradient checkpointing, StreamBP scales up the maximum sequence length of BP by 2.8-5.5 times larger, while using comparable or even less BP time. Note that StreamBP's sequence length scaling ability can be directly transferred to batch size scaling for accelerating training. We further develop a communication-efficient distributed StreamBP to effectively support multi-GPU training and broaden its applicability. Our code can be easily integrated into the training pipeline of any transformer models and is available at https://github.com/Ledzy/StreamBP. |
2025-06-06T00:00:00 | 2506.05331 | MINT-CoT: Enabling Interleaved Visual Tokens in Mathematical Chain-of-Thought Reasoning | [
"Xinyan Chen",
"Renrui Zhang",
"Dongzhi Jiang",
"Aojun Zhou",
"Shilin Yan",
"Weifeng Lin",
"Hongsheng Li"
] | https://github.com/xinyan-cxy/MINT-CoT | Chain-of-Thought (CoT) has widely enhanced mathematical reasoning in Large Language Models (LLMs), but it still remains challenging for extending it to multimodal domains. Existing works either adopt a similar textual reasoning for image input, or seek to interleave visual signals into mathematical CoT. However, they face three key limitations for math problem-solving: reliance on coarse-grained box-shaped image regions, limited perception of vision encoders on math content, and dependence on external capabilities for visual modification. In this paper, we propose MINT-CoT, introducing Mathematical INterleaved Tokens for Chain-of-Thought visual reasoning. MINT-CoT adaptively interleaves relevant visual tokens into textual reasoning steps via an Interleave Token, which dynamically selects visual regions of any shapes within math figures. To empower this capability, we construct the MINT-CoT dataset, containing 54K mathematical problems aligning each reasoning step with visual regions at the token level, accompanied by a rigorous data generation pipeline. We further present a three-stage MINT-CoT training strategy, progressively combining text-only CoT SFT, interleaved CoT SFT, and interleaved CoT RL, which derives our MINT-CoT-7B model. Extensive experiments demonstrate the effectiveness of our method for effective visual interleaved reasoning in mathematical domains, where MINT-CoT-7B outperforms the baseline model by +34.08% on MathVista, +28.78% on GeoQA, and +23.2% on MMStar, respectively. Our code and data are available at https://github.com/xinyan-cxy/MINT-CoT |
2025-06-06T00:00:00 | 2506.04633 | Unfolding Spatial Cognition: Evaluating Multimodal Models on Visual Simulations | [
"Linjie Li",
"Mahtab Bigverdi",
"Jiawei Gu",
"Zixian Ma",
"Yinuo Yang",
"Ziang Li",
"Yejin Choi",
"Ranjay Krishna"
] | Spatial cognition is essential for human intelligence, enabling problem-solving through visual simulations rather than solely relying on verbal reasoning. However, existing AI benchmarks primarily assess verbal reasoning, neglecting the complexities of non-verbal, multi-step visual simulation. We introduce STARE(Spatial Transformations and Reasoning Evaluation), a benchmark designed to rigorously evaluate multimodal large language models on tasks better solved through multi-step visual simulation. STARE features 4K tasks spanning foundational geometric transformations (2D and 3D), integrated spatial reasoning (cube net folding and tangram puzzles), and real-world spatial reasoning (perspective and temporal reasoning), reflecting practical cognitive challenges like object assembly, mechanical diagram interpretation, and everyday spatial navigation. Our evaluations show that models excel at reasoning over simpler 2D transformations, but perform close to random chance on more complex tasks like 3D cube net folding and tangram puzzles that require multi-step visual simulations. Humans achieve near-perfect accuracy but take considerable time (up to 28.9s) on complex tasks, significantly speeding up (down by 7.5 seconds on average) with intermediate visual simulations. In contrast, models exhibit inconsistent performance gains from visual simulations, improving on most tasks but declining in specific cases like tangram puzzles (GPT-4o, o1) and cube net folding (Claude-3.5, Gemini-2.0 Flash), indicating that models may not know how to effectively leverage intermediate visual information. |
|
2025-06-06T00:00:00 | 2505.20914 | Geometry-Editable and Appearance-Preserving Object Compositon | [
"Jianman Lin",
"Haojie Li",
"Chunmei Qing",
"Zhijing Yang",
"Liang Lin",
"Tianshui Chen"
] | General object composition (GOC) aims to seamlessly integrate a target object into a background scene with desired geometric properties, while simultaneously preserving its fine-grained appearance details. Recent approaches derive semantic embeddings and integrate them into advanced diffusion models to enable geometry-editable generation. However, these highly compact embeddings encode only high-level semantic cues and inevitably discard fine-grained appearance details. We introduce a Disentangled Geometry-editable and Appearance-preserving Diffusion (DGAD) model that first leverages semantic embeddings to implicitly capture the desired geometric transformations and then employs a cross-attention retrieval mechanism to align fine-grained appearance features with the geometry-edited representation, facilitating both precise geometry editing and faithful appearance preservation in object composition. Specifically, DGAD builds on CLIP/DINO-derived and reference networks to extract semantic embeddings and appearance-preserving representations, which are then seamlessly integrated into the encoding and decoding pipelines in a disentangled manner. We first integrate the semantic embeddings into pre-trained diffusion models that exhibit strong spatial reasoning capabilities to implicitly capture object geometry, thereby facilitating flexible object manipulation and ensuring effective editability. Then, we design a dense cross-attention mechanism that leverages the implicitly learned object geometry to retrieve and spatially align appearance features with their corresponding regions, ensuring faithful appearance consistency. Extensive experiments on public benchmarks demonstrate the effectiveness of the proposed DGAD framework. |
|
2025-06-06T00:00:00 | 2506.05282 | Rectified Point Flow: Generic Point Cloud Pose Estimation | [
"Tao Sun",
"Liyuan Zhu",
"Shengyu Huang",
"Shuran Song",
"Iro Armeni"
] | We introduce Rectified Point Flow, a unified parameterization that formulates pairwise point cloud registration and multi-part shape assembly as a single conditional generative problem. Given unposed point clouds, our method learns a continuous point-wise velocity field that transports noisy points toward their target positions, from which part poses are recovered. In contrast to prior work that regresses part-wise poses with ad-hoc symmetry handling, our method intrinsically learns assembly symmetries without symmetry labels. Together with a self-supervised encoder focused on overlapping points, our method achieves a new state-of-the-art performance on six benchmarks spanning pairwise registration and shape assembly. Notably, our unified formulation enables effective joint training on diverse datasets, facilitating the learning of shared geometric priors and consequently boosting accuracy. Project page: https://rectified-pointflow.github.io/. |
|
2025-06-06T00:00:00 | 2506.05010 | ComfyUI-Copilot: An Intelligent Assistant for Automated Workflow Development | [
"Zhenran Xu",
"Xue Yang",
"Yiyu Wang",
"Qingli Hu",
"Zijiao Wu",
"Longyue Wang",
"Weihua Luo",
"Kaifu Zhang",
"Baotian Hu",
"Min Zhang"
] | https://github.com/AIDC-AI/ComfyUI-Copilot | We introduce ComfyUI-Copilot, a large language model-powered plugin designed to enhance the usability and efficiency of ComfyUI, an open-source platform for AI-driven art creation. Despite its flexibility and user-friendly interface, ComfyUI can present challenges to newcomers, including limited documentation, model misconfigurations, and the complexity of workflow design. ComfyUI-Copilot addresses these challenges by offering intelligent node and model recommendations, along with automated one-click workflow construction. At its core, the system employs a hierarchical multi-agent framework comprising a central assistant agent for task delegation and specialized worker agents for different usages, supported by our curated ComfyUI knowledge bases to streamline debugging and deployment. We validate the effectiveness of ComfyUI-Copilot through both offline quantitative evaluations and online user feedback, showing that it accurately recommends nodes and accelerates workflow development. Additionally, use cases illustrate that ComfyUI-Copilot lowers entry barriers for beginners and enhances workflow efficiency for experienced users. The ComfyUI-Copilot installation package and a demo video are available at https://github.com/AIDC-AI/ComfyUI-Copilot. |
2025-06-06T00:00:00 | 2506.02620 | FlexPainter: Flexible and Multi-View Consistent Texture Generation | [
"Dongyu Yan",
"Leyi Wu",
"Jiantao Lin",
"Luozhou Wang",
"Tianshuo Xu",
"Zhifei Chen",
"Zhen Yang",
"Lie Xu",
"Shunsi Zhang",
"Yingcong Chen"
] | Texture map production is an important part of 3D modeling and determines the rendering quality. Recently, diffusion-based methods have opened a new way for texture generation. However, restricted control flexibility and limited prompt modalities may prevent creators from producing desired results. Furthermore, inconsistencies between generated multi-view images often lead to poor texture generation quality. To address these issues, we introduce FlexPainter, a novel texture generation pipeline that enables flexible multi-modal conditional guidance and achieves highly consistent texture generation. A shared conditional embedding space is constructed to perform flexible aggregation between different input modalities. Utilizing such embedding space, we present an image-based CFG method to decompose structural and style information, achieving reference image-based stylization. Leveraging the 3D knowledge within the image diffusion prior, we first generate multi-view images simultaneously using a grid representation to enhance global understanding. Meanwhile, we propose a view synchronization and adaptive weighting module during diffusion sampling to further ensure local consistency. Finally, a 3D-aware texture completion model combined with a texture enhancement model is used to generate seamless, high-resolution texture maps. Comprehensive experiments demonstrate that our framework significantly outperforms state-of-the-art methods in both flexibility and generation quality. |
|
2025-06-06T00:00:00 | 2506.00830 | SkyReels-Audio: Omni Audio-Conditioned Talking Portraits in Video Diffusion Transformers | [
"Zhengcong Fei",
"Hao Jiang",
"Di Qiu",
"Baoxuan Gu",
"Youqiang Zhang",
"Jiahua Wang",
"Jialin Bai",
"Debang Li",
"Mingyuan Fan",
"Guibin Chen",
"Yahui Zhou"
] | The generation and editing of audio-conditioned talking portraits guided by multimodal inputs, including text, images, and videos, remains under explored. In this paper, we present SkyReels-Audio, a unified framework for synthesizing high-fidelity and temporally coherent talking portrait videos. Built upon pretrained video diffusion transformers, our framework supports infinite-length generation and editing, while enabling diverse and controllable conditioning through multimodal inputs. We employ a hybrid curriculum learning strategy to progressively align audio with facial motion, enabling fine-grained multimodal control over long video sequences. To enhance local facial coherence, we introduce a facial mask loss and an audio-guided classifier-free guidance mechanism. A sliding-window denoising approach further fuses latent representations across temporal segments, ensuring visual fidelity and temporal consistency across extended durations and diverse identities. More importantly, we construct a dedicated data pipeline for curating high-quality triplets consisting of synchronized audio, video, and textual descriptions. Comprehensive benchmark evaluations show that SkyReels-Audio achieves superior performance in lip-sync accuracy, identity consistency, and realistic facial dynamics, particularly under complex and challenging conditions. |
|
2025-06-06T00:00:00 | 2506.05278 | Micro-Act: Mitigate Knowledge Conflict in Question Answering via Actionable Self-Reasoning | [
"Nan Huo",
"Jinyang Li",
"Bowen Qin",
"Ge Qu",
"Xiaolong Li",
"Xiaodong Li",
"Chenhao Ma",
"Reynold Cheng"
] | Retrieval-Augmented Generation (RAG) systems commonly suffer from Knowledge Conflicts, where retrieved external knowledge contradicts the inherent, parametric knowledge of large language models (LLMs). It adversely affects performance on downstream tasks such as question answering (QA). Existing approaches often attempt to mitigate conflicts by directly comparing two knowledge sources in a side-by-side manner, but this can overwhelm LLMs with extraneous or lengthy contexts, ultimately hindering their ability to identify and mitigate inconsistencies. To address this issue, we propose Micro-Act a framework with a hierarchical action space that automatically perceives context complexity and adaptively decomposes each knowledge source into a sequence of fine-grained comparisons. These comparisons are represented as actionable steps, enabling reasoning beyond the superficial context. Through extensive experiments on five benchmark datasets, Micro-Act consistently achieves significant increase in QA accuracy over state-of-the-art baselines across all 5 datasets and 3 conflict types, especially in temporal and semantic types where all baselines fail significantly. More importantly, Micro-Act exhibits robust performance on non-conflict questions simultaneously, highlighting its practical value in real-world RAG applications. |
|
2025-06-06T00:00:00 | 2506.01011 | Autoregressive Images Watermarking through Lexical Biasing: An Approach Resistant to Regeneration Attack | [
"Siqi Hui",
"Yiren Song",
"Sanping Zhou",
"Ye Deng",
"Wenli Huang",
"Jinjun Wang"
] | Autoregressive (AR) image generation models have gained increasing attention for their breakthroughs in synthesis quality, highlighting the need for robust watermarking to prevent misuse. However, existing in-generation watermarking techniques are primarily designed for diffusion models, where watermarks are embedded within diffusion latent states. This design poses significant challenges for direct adaptation to AR models, which generate images sequentially through token prediction. Moreover, diffusion-based regeneration attacks can effectively erase such watermarks by perturbing diffusion latent states. To address these challenges, we propose Lexical Bias Watermarking (LBW), a novel framework designed for AR models that resists regeneration attacks. LBW embeds watermarks directly into token maps by biasing token selection toward a predefined green list during generation. This approach ensures seamless integration with existing AR models and extends naturally to post-hoc watermarking. To increase the security against white-box attacks, instead of using a single green list, the green list for each image is randomly sampled from a pool of green lists. Watermark detection is performed via quantization and statistical analysis of the token distribution. Extensive experiments demonstrate that LBW achieves superior watermark robustness, particularly in resisting regeneration attacks. |
|
2025-06-06T00:00:00 | 2506.05349 | VideoMathQA: Benchmarking Mathematical Reasoning via Multimodal Understanding in Videos | [
"Hanoona Rasheed",
"Abdelrahman Shaker",
"Anqi Tang",
"Muhammad Maaz",
"Ming-Hsuan Yang",
"Salman Khan",
"Fahad Khan"
] | Mathematical reasoning in real-world video settings presents a fundamentally different challenge than in static images or text. It requires interpreting fine-grained visual information, accurately reading handwritten or digital text, and integrating spoken cues, often dispersed non-linearly over time. In such multimodal contexts, success hinges not just on perception, but on selectively identifying and integrating the right contextual details from a rich and noisy stream of content. To this end, we introduce VideoMathQA, a benchmark designed to evaluate whether models can perform such temporally extended cross-modal reasoning on videos. The benchmark spans 10 diverse mathematical domains, covering videos ranging from 10 seconds to over 1 hour. It requires models to interpret structured visual content, understand instructional narratives, and jointly ground concepts across visual, audio, and textual modalities. We employ graduate-level experts to ensure high quality, totaling over 920 man-hours of annotation. To reflect real-world scenarios, questions are designed around three core reasoning challenges: direct problem solving, where answers are grounded in the presented question; conceptual transfer, which requires applying learned methods to new problems; and deep instructional comprehension, involving multi-step reasoning over extended explanations and partially worked-out solutions. Each question includes multi-step reasoning annotations, enabling fine-grained diagnosis of model capabilities. Through this benchmark, we highlight the limitations of existing approaches and establish a systematic evaluation framework for models that must reason, rather than merely perceive, across temporally extended and modality-rich mathematical problem settings. Our benchmark and evaluation code are available at: https://mbzuai-oryx.github.io/VideoMathQA |
|
2025-06-06T00:00:00 | 2506.02587 | BEVCALIB: LiDAR-Camera Calibration via Geometry-Guided Bird's-Eye View Representations | [
"Weiduo Yuan",
"Jerry Li",
"Justin Yue",
"Divyank Shah",
"Konstantinos Karydis",
"Hang Qiu"
] | Accurate LiDAR-camera calibration is fundamental to fusing multi-modal perception in autonomous driving and robotic systems. Traditional calibration methods require extensive data collection in controlled environments and cannot compensate for the transformation changes during the vehicle/robot movement. In this paper, we propose the first model that uses bird's-eye view (BEV) features to perform LiDAR camera calibration from raw data, termed BEVCALIB. To achieve this, we extract camera BEV features and LiDAR BEV features separately and fuse them into a shared BEV feature space. To fully utilize the geometric information from the BEV feature, we introduce a novel feature selector to filter the most important features in the transformation decoder, which reduces memory consumption and enables efficient training. Extensive evaluations on KITTI, NuScenes, and our own dataset demonstrate that BEVCALIB establishes a new state of the art. Under various noise conditions, BEVCALIB outperforms the best baseline in the literature by an average of (47.08%, 82.32%) on KITTI dataset, and (78.17%, 68.29%) on NuScenes dataset, in terms of (translation, rotation), respectively. In the open-source domain, it improves the best reproducible baseline by one order of magnitude. Our code and demo results are available at https://cisl.ucr.edu/BEVCalib. |
|
2025-06-06T00:00:00 | 2506.05284 | Video World Models with Long-term Spatial Memory | [
"Tong Wu",
"Shuai Yang",
"Ryan Po",
"Yinghao Xu",
"Ziwei Liu",
"Dahua Lin",
"Gordon Wetzstein"
] | Emerging world models autoregressively generate video frames in response to actions, such as camera movements and text prompts, among other control signals. Due to limited temporal context window sizes, these models often struggle to maintain scene consistency during revisits, leading to severe forgetting of previously generated environments. Inspired by the mechanisms of human memory, we introduce a novel framework to enhancing long-term consistency of video world models through a geometry-grounded long-term spatial memory. Our framework includes mechanisms to store and retrieve information from the long-term spatial memory and we curate custom datasets to train and evaluate world models with explicitly stored 3D memory mechanisms. Our evaluations show improved quality, consistency, and context length compared to relevant baselines, paving the way towards long-term consistent world generation. |
|
2025-06-06T00:00:00 | 2506.05209 | The Common Pile v0.1: An 8TB Dataset of Public Domain and Openly Licensed Text | [
"Nikhil Kandpal",
"Brian Lester",
"Colin Raffel",
"Sebastian Majstorovic",
"Stella Biderman",
"Baber Abbasi",
"Luca Soldaini",
"Enrico Shippole",
"A. Feder Cooper",
"Aviya Skowron",
"John Kirchenbauer",
"Shayne Longpre",
"Lintang Sutawika",
"Alon Albalak",
"Zhenlin Xu",
"Guilherme Penedo",
"Loubna Ben Allal",
"Elie Bakouch",
"John David Pressman",
"Honglu Fan",
"Dashiell Stander",
"Guangyu Song",
"Aaron Gokaslan",
"Tom Goldstein",
"Brian R. Bartoldson",
"Bhavya Kailkhura",
"Tyler Murray"
] | Large language models (LLMs) are typically trained on enormous quantities of unlicensed text, a practice that has led to scrutiny due to possible intellectual property infringement and ethical concerns. Training LLMs on openly licensed text presents a first step towards addressing these issues, but prior data collection efforts have yielded datasets too small or low-quality to produce performant LLMs. To address this gap, we collect, curate, and release the Common Pile v0.1, an eight terabyte collection of openly licensed text designed for LLM pretraining. The Common Pile comprises content from 30 sources that span diverse domains including research papers, code, books, encyclopedias, educational materials, audio transcripts, and more. Crucially, we validate our efforts by training two 7 billion parameter LLMs on text from the Common Pile: Comma v0.1-1T and Comma v0.1-2T, trained on 1 and 2 trillion tokens respectively. Both models attain competitive performance to LLMs trained on unlicensed text with similar computational budgets, such as Llama 1 and 2 7B. In addition to releasing the Common Pile v0.1 itself, we also release the code used in its creation as well as the training mixture and checkpoints for the Comma v0.1 models. |
|
2025-06-06T00:00:00 | 2506.04996 | PATS: Proficiency-Aware Temporal Sampling for Multi-View Sports Skill Assessment | [
"Edoardo Bianchi",
"Antonio Liotta"
] | Automated sports skill assessment requires capturing fundamental movement patterns that distinguish expert from novice performance, yet current video sampling methods disrupt the temporal continuity essential for proficiency evaluation. To this end, we introduce Proficiency-Aware Temporal Sampling (PATS), a novel sampling strategy that preserves complete fundamental movements within continuous temporal segments for multi-view skill assessment. PATS adaptively segments videos to ensure each analyzed portion contains full execution of critical performance components, repeating this process across multiple segments to maximize information coverage while maintaining temporal coherence. Evaluated on the EgoExo4D benchmark with SkillFormer, PATS surpasses the state-of-the-art accuracy across all viewing configurations (+0.65% to +3.05%) and delivers substantial gains in challenging domains (+26.22% bouldering, +2.39% music, +1.13% basketball). Systematic analysis reveals that PATS successfully adapts to diverse activity characteristics-from high-frequency sampling for dynamic sports to fine-grained segmentation for sequential skills-demonstrating its effectiveness as an adaptive approach to temporal sampling that advances automated skill assessment for real-world applications. |
|
2025-06-06T00:00:00 | 2506.02865 | Surfer-H Meets Holo1: Cost-Efficient Web Agent Powered by Open Weights | [
"Mathieu Andreux",
"Breno Baldas Skuk",
"Hamza Benchekroun",
"Emilien Biré",
"Antoine Bonnet",
"Riaz Bordie",
"Matthias Brunel",
"Pierre-Louis Cedoz",
"Antoine Chassang",
"Mickaël Chen",
"Alexandra D. Constantinou",
"Antoine d'Andigné",
"Hubert de La Jonquière",
"Aurélien Delfosse",
"Ludovic Denoyer",
"Alexis Deprez",
"Augustin Derupti",
"Michael Eickenberg",
"Mathïs Federico",
"Charles Kantor",
"Xavier Koegler",
"Yann Labbé",
"Matthew C. H. Lee",
"Erwan Le Jumeau de Kergaradec",
"Amir Mahla",
"Avshalom Manevich",
"Adrien Maret",
"Charles Masson",
"Rafaël Maurin",
"Arturo Mena",
"Philippe Modard",
"Axel Moyal",
"Axel Nguyen Kerbel",
"Julien Revelle",
"Mats L. Richter",
"María Santos",
"Laurent Sifre",
"Maxime Theillard",
"Marc Thibault",
"Louis Thiry",
"Léo Tronchon",
"Nicolas Usunier",
"Tony Wu"
] | We present Surfer-H, a cost-efficient web agent that integrates Vision-Language Models (VLM) to perform user-defined tasks on the web. We pair it with Holo1, a new open-weight collection of VLMs specialized in web navigation and information extraction. Holo1 was trained on carefully curated data sources, including open-access web content, synthetic examples, and self-produced agentic data. Holo1 tops generalist User Interface (UI) benchmarks as well as our new web UI localization benchmark, WebClick. When powered by Holo1, Surfer-H achieves a 92.2% state-of-the-art performance on WebVoyager, striking a Pareto-optimal balance between accuracy and cost-efficiency. To accelerate research advancement in agentic systems, we are open-sourcing both our WebClick evaluation dataset and the Holo1 model weights. |
|
2025-06-06T00:00:00 | 2506.05301 | SeedVR2: One-Step Video Restoration via Diffusion Adversarial Post-Training | [
"Jianyi Wang",
"Shanchuan Lin",
"Zhijie Lin",
"Yuxi Ren",
"Meng Wei",
"Zongsheng Yue",
"Shangchen Zhou",
"Hao Chen",
"Yang Zhao",
"Ceyuan Yang",
"Xuefeng Xiao",
"Chen Change Loy",
"Lu Jiang"
] | Recent advances in diffusion-based video restoration (VR) demonstrate significant improvement in visual quality, yet yield a prohibitive computational cost during inference. While several distillation-based approaches have exhibited the potential of one-step image restoration, extending existing approaches to VR remains challenging and underexplored, particularly when dealing with high-resolution video in real-world settings. In this work, we propose a one-step diffusion-based VR model, termed as SeedVR2, which performs adversarial VR training against real data. To handle the challenging high-resolution VR within a single step, we introduce several enhancements to both model architecture and training procedures. Specifically, an adaptive window attention mechanism is proposed, where the window size is dynamically adjusted to fit the output resolutions, avoiding window inconsistency observed under high-resolution VR using window attention with a predefined window size. To stabilize and improve the adversarial post-training towards VR, we further verify the effectiveness of a series of losses, including a proposed feature matching loss without significantly sacrificing training efficiency. Extensive experiments show that SeedVR2 can achieve comparable or even better performance compared with existing VR approaches in a single step. |
|
2025-06-06T00:00:00 | 2506.05345 | Inference-Time Hyper-Scaling with KV Cache Compression | [
"Adrian Łańcucki",
"Konrad Staniszewski",
"Piotr Nawrot",
"Edoardo M. Ponti"
] | Inference-time scaling trades efficiency for increased reasoning accuracy by generating longer or more parallel sequences. However, in Transformer LLMs, generation cost is bottlenecked by the size of the key-value (KV) cache, rather than the number of generated tokens. Hence, we explore inference-time hyper-scaling: by compressing the KV cache, we can generate more tokens within the same compute budget and further improve the accuracy of scaled inference. The success of this approach, however, hinges on the ability of compression methods to preserve accuracy even at high compression ratios. To make hyper-scaling practical, we introduce Dynamic Memory Sparsification (DMS), a novel method for sparsifying KV caches that only requires 1K training steps to achieve 8times compression, while maintaining better accuracy than training-free sparse attention. Instead of prematurely discarding cached tokens, DMS delays token eviction, implicitly merging representations and preserving critical information. We demonstrate the effectiveness of inference-time hyper-scaling with DMS on multiple families of LLMs, showing that it boosts accuracy for comparable inference runtime and memory load. For instance, we enhance Qwen-R1 32B by an average of 9.1 points on AIME 24, 7.6 on GPQA, and 9.6 on LiveCodeBench across compute budgets. |
|
2025-06-06T00:00:00 | 2506.02751 | RobustSplat: Decoupling Densification and Dynamics for Transient-Free 3DGS | [
"Chuanyu Fu",
"Yuqi Zhang",
"Kunbin Yao",
"Guanying Chen",
"Yuan Xiong",
"Chuan Huang",
"Shuguang Cui",
"Xiaochun Cao"
] | 3D Gaussian Splatting (3DGS) has gained significant attention for its real-time, photo-realistic rendering in novel-view synthesis and 3D modeling. However, existing methods struggle with accurately modeling scenes affected by transient objects, leading to artifacts in the rendered images. We identify that the Gaussian densification process, while enhancing scene detail capture, unintentionally contributes to these artifacts by growing additional Gaussians that model transient disturbances. To address this, we propose RobustSplat, a robust solution based on two critical designs. First, we introduce a delayed Gaussian growth strategy that prioritizes optimizing static scene structure before allowing Gaussian splitting/cloning, mitigating overfitting to transient objects in early optimization. Second, we design a scale-cascaded mask bootstrapping approach that first leverages lower-resolution feature similarity supervision for reliable initial transient mask estimation, taking advantage of its stronger semantic consistency and robustness to noise, and then progresses to high-resolution supervision to achieve more precise mask prediction. Extensive experiments on multiple challenging datasets show that our method outperforms existing methods, clearly demonstrating the robustness and effectiveness of our method. Our project page is https://fcyycf.github.io/RobustSplat/. |
|
2025-06-06T00:00:00 | 2506.05348 | FreeTimeGS: Free Gaussians at Anytime and Anywhere for Dynamic Scene Reconstruction | [
"Yifan Wang",
"Peishan Yang",
"Zhen Xu",
"Jiaming Sun",
"Zhanhua Zhang",
"Yong Chen",
"Hujun Bao",
"Sida Peng",
"Xiaowei Zhou"
] | This paper addresses the challenge of reconstructing dynamic 3D scenes with complex motions. Some recent works define 3D Gaussian primitives in the canonical space and use deformation fields to map canonical primitives to observation spaces, achieving real-time dynamic view synthesis. However, these methods often struggle to handle scenes with complex motions due to the difficulty of optimizing deformation fields. To overcome this problem, we propose FreeTimeGS, a novel 4D representation that allows Gaussian primitives to appear at arbitrary time and locations. In contrast to canonical Gaussian primitives, our representation possesses the strong flexibility, thus improving the ability to model dynamic 3D scenes. In addition, we endow each Gaussian primitive with an motion function, allowing it to move to neighboring regions over time, which reduces the temporal redundancy. Experiments results on several datasets show that the rendering quality of our method outperforms recent methods by a large margin. |
|
2025-06-06T00:00:00 | 2505.23115 | Diffusion-Based Generative Models for 3D Occupancy Prediction in Autonomous Driving | [
"Yunshen Wang",
"Yicheng Liu",
"Tianyuan Yuan",
"Yucheng Mao",
"Yingshi Liang",
"Xiuyu Yang",
"Honggang Zhang",
"Hang Zhao"
] | Accurately predicting 3D occupancy grids from visual inputs is critical for autonomous driving, but current discriminative methods struggle with noisy data, incomplete observations, and the complex structures inherent in 3D scenes. In this work, we reframe 3D occupancy prediction as a generative modeling task using diffusion models, which learn the underlying data distribution and incorporate 3D scene priors. This approach enhances prediction consistency, noise robustness, and better handles the intricacies of 3D spatial structures. Our extensive experiments show that diffusion-based generative models outperform state-of-the-art discriminative approaches, delivering more realistic and accurate occupancy predictions, especially in occluded or low-visibility regions. Moreover, the improved predictions significantly benefit downstream planning tasks, highlighting the practical advantages of our method for real-world autonomous driving applications. |
|
2025-06-06T00:00:00 | 2506.05229 | Diagonal Batching Unlocks Parallelism in Recurrent Memory Transformers for Long Contexts | [
"Danil Sivtsov",
"Ivan Rodkin",
"Gleb Kuzmin",
"Yuri Kuratov",
"Ivan Oseledets"
] | Transformer models struggle with long-context inference due to their quadratic time and linear memory complexity. Recurrent Memory Transformers (RMTs) offer a solution by reducing the asymptotic cost to linear time and constant memory usage. However, their memory update mechanism leads to sequential execution, causing a performance bottleneck. We introduce Diagonal Batching, a scheduling scheme that unlocks parallelism across segments in RMTs while preserving exact recurrence. This approach eliminates the sequential constraint, enabling efficient GPU inference even for single long-context inputs without complex batching and pipelining techniques. Because the technique is purely a run-time computation reordering, existing RMT models adopt it with no retraining. Applied to a LLaMA-1B ARMT model, Diagonal Batching yields a 3.3x speedup over standard full-attention LLaMA-1B and a 1.8x speedup over the sequential RMT implementation on 131,072-token sequences. By removing sequential bottleneck, Diagonal Batching reduces inference cost and latency, thereby strengthening RMTs as a practical solution for real-world, long-context applications. |
|
2025-06-06T00:00:00 | 2506.00981 | What do self-supervised speech models know about Dutch? Analyzing advantages of language-specific pre-training | [
"Marianne de Heer Kloots",
"Hosein Mohebbi",
"Charlotte Pouw",
"Gaofei Shen",
"Willem Zuidema",
"Martijn Bentum"
] | How language-specific are speech representations learned by self-supervised models? Existing work has shown that a range of linguistic features can be successfully decoded from end-to-end models trained only on speech recordings. However, it's less clear to what extent pre-training on specific languages improves language-specific linguistic information. Here we test the encoding of Dutch phonetic and lexical information in internal representations of self-supervised Wav2Vec2 models. Pre-training exclusively on Dutch improves the representation of Dutch linguistic features as compared to pre-training on similar amounts of English or larger amounts of multilingual data. This language-specific advantage is well-detected by trained clustering or classification probes, and partially observable using zero-shot metrics. Furthermore, the language-specific benefit on linguistic feature encoding aligns with downstream performance on Automatic Speech Recognition. |
|
2025-06-06T00:00:00 | 2506.04956 | FEAT: Full-Dimensional Efficient Attention Transformer for Medical Video Generation | [
"Huihan Wang",
"Zhiwen Yang",
"Hui Zhang",
"Dan Zhao",
"Bingzheng Wei",
"Yan Xu"
] | https://github.com/Yaziwel/FEAT | Synthesizing high-quality dynamic medical videos remains a significant challenge due to the need for modeling both spatial consistency and temporal dynamics. Existing Transformer-based approaches face critical limitations, including insufficient channel interactions, high computational complexity from self-attention, and coarse denoising guidance from timestep embeddings when handling varying noise levels. In this work, we propose FEAT, a full-dimensional efficient attention Transformer, which addresses these issues through three key innovations: (1) a unified paradigm with sequential spatial-temporal-channel attention mechanisms to capture global dependencies across all dimensions, (2) a linear-complexity design for attention mechanisms in each dimension, utilizing weighted key-value attention and global channel attention, and (3) a residual value guidance module that provides fine-grained pixel-level guidance to adapt to different noise levels. We evaluate FEAT on standard benchmarks and downstream tasks, demonstrating that FEAT-S, with only 23\% of the parameters of the state-of-the-art model Endora, achieves comparable or even superior performance. Furthermore, FEAT-L surpasses all comparison methods across multiple datasets, showcasing both superior effectiveness and scalability. Code is available at https://github.com/Yaziwel/FEAT. |
2025-06-06T00:00:00 | 2506.04598 | Scaling Laws for Robust Comparison of Open Foundation Language-Vision Models and Datasets | [
"Marianna Nezhurina",
"Tomer Porian",
"Giovanni Pucceti",
"Tommie Kerssies",
"Romain Beaumont",
"Mehdi Cherti",
"Jenia Jitsev"
] | https://github.com/LAION-AI/scaling-laws-for-comparison | In studies of transferable learning, scaling laws are obtained for various important foundation models to predict their properties and performance at larger scales. We show here how scaling law derivation can also be used for model and dataset comparison, allowing to decide which procedure is to be preferred for pre-training. For the first time, full scaling laws based on dense measurements across a wide span of model and samples seen scales are derived for two important language-vision learning procedures, CLIP and MaMMUT, that use either contrastive only or contrastive and captioning text generative loss. Ensuring sufficient prediction accuracy for held out points, we use derived scaling laws to compare both models, obtaining evidence for MaMMUT's stronger improvement with scale and better sample efficiency than standard CLIP. To strengthen validity of the comparison, we show scaling laws for various downstream tasks, classification, retrieval, and segmentation, and for different open datasets, DataComp, DFN and Re-LAION, observing consistently the same trends. We show that comparison can also be performed when deriving scaling laws with a constant learning rate schedule, reducing compute cost. Accurate derivation of scaling laws provides thus means to perform model and dataset comparison across scale spans, avoiding misleading conclusions based on measurements from single reference scales only, paving the road for systematic comparison and improvement of open foundation models and datasets for their creation. We release all the pre-trained models with their intermediate checkpoints, including openMaMMUT-L/14, which achieves 80.3% zero-shot ImageNet-1k accuracy, trained on 12.8B samples from DataComp-1.4B. Code for reproducing experiments in the paper and raw experiments data can be found at https://github.com/LAION-AI/scaling-laws-for-comparison. |
2025-06-06T00:00:00 | 2506.03643 | Images are Worth Variable Length of Representations | [
"Lingjun Mao",
"Rodolfo Corona",
"Xin Liang",
"Wenhao Yan",
"Zineng Tang"
] | Most existing vision encoders map images into a fixed-length sequence of tokens, overlooking the fact that different images contain varying amounts of information. For example, a visually complex image (e.g., a cluttered room) inherently carries more information and thus deserves more tokens than a simple image (e.g., a blank wall). To address this inefficiency, we propose DOVE, a dynamic vision encoder that produces a variable number of visual tokens (i.e., continuous representation vectors) to reconstruct each image. Our results show that DOVE significantly reduces the average number of tokens while maintaining high reconstruction quality. In several linear probing and downstream multimodal tasks, it outperforms existing autoencoder-based tokenization methods when using far fewer tokens, capturing more expressive semantic features compared to fixed-length encoding. We further extend DOVE with query-conditioned tokenization. By guiding the model to focus on query-relevant regions, it achieves more efficient and targeted semantic extraction. Our code and checkpoints are available at https://dove-encoder.github.io/dove-encoder. |
|
2025-06-06T00:00:00 | 2506.02444 | SViMo: Synchronized Diffusion for Video and Motion Generation in Hand-object Interaction Scenarios | [
"Lingwei Dang",
"Ruizhi Shao",
"Hongwen Zhang",
"Wei Min",
"Yebin Liu",
"Qingyao Wu"
] | https://github.com/Droliven/SViMo\_project | Hand-Object Interaction (HOI) generation has significant application potential. However, current 3D HOI motion generation approaches heavily rely on predefined 3D object models and lab-captured motion data, limiting generalization capabilities. Meanwhile, HOI video generation methods prioritize pixel-level visual fidelity, often sacrificing physical plausibility. Recognizing that visual appearance and motion patterns share fundamental physical laws in the real world, we propose a novel framework that combines visual priors and dynamic constraints within a synchronized diffusion process to generate the HOI video and motion simultaneously. To integrate the heterogeneous semantics, appearance, and motion features, our method implements tri-modal adaptive modulation for feature aligning, coupled with 3D full-attention for modeling inter- and intra-modal dependencies. Furthermore, we introduce a vision-aware 3D interaction diffusion model that generates explicit 3D interaction sequences directly from the synchronized diffusion outputs, then feeds them back to establish a closed-loop feedback cycle. This architecture eliminates dependencies on predefined object models or explicit pose guidance while significantly enhancing video-motion consistency. Experimental results demonstrate our method's superiority over state-of-the-art approaches in generating high-fidelity, dynamically plausible HOI sequences, with notable generalization capabilities in unseen real-world scenarios. Project page at https://github.com/Droliven/SViMo\_project. |
2025-06-06T00:00:00 | 2506.04462 | Watermarking Degrades Alignment in Language Models: Analysis and Mitigation | [
"Apurv Verma",
"NhatHai Phan",
"Shubhendu Trivedi"
] | Watermarking techniques for large language models (LLMs) can significantly impact output quality, yet their effects on truthfulness, safety, and helpfulness remain critically underexamined. This paper presents a systematic analysis of how two popular watermarking approaches-Gumbel and KGW-affect these core alignment properties across four aligned LLMs. Our experiments reveal two distinct degradation patterns: guard attenuation, where enhanced helpfulness undermines model safety, and guard amplification, where excessive caution reduces model helpfulness. These patterns emerge from watermark-induced shifts in token distribution, surfacing the fundamental tension that exists between alignment objectives. To mitigate these degradations, we propose Alignment Resampling (AR), an inference-time sampling method that uses an external reward model to restore alignment. We establish a theoretical lower bound on the improvement in expected reward score as the sample size is increased and empirically demonstrate that sampling just 2-4 watermarked generations effectively recovers or surpasses baseline (unwatermarked) alignment scores. To overcome the limited response diversity of standard Gumbel watermarking, our modified implementation sacrifices strict distortion-freeness while maintaining robust detectability, ensuring compatibility with AR. Experimental results confirm that AR successfully recovers baseline alignment in both watermarking approaches, while maintaining strong watermark detectability. This work reveals the critical balance between watermark strength and model alignment, providing a simple inference-time solution to responsibly deploy watermarked LLMs in practice. |
|
2025-06-06T00:00:00 | 2506.05334 | Search Arena: Analyzing Search-Augmented LLMs | [
"Mihran Miroyan",
"Tsung-Han Wu",
"Logan King",
"Tianle Li",
"Jiayi Pan",
"Xinyan Hu",
"Wei-Lin Chiang",
"Anastasios N. Angelopoulos",
"Trevor Darrell",
"Narges Norouzi",
"Joseph E. Gonzalez"
] | https://github.com/lmarena/search-arena | Search-augmented language models combine web search with Large Language Models (LLMs) to improve response groundedness and freshness. However, analyzing these systems remains challenging: existing datasets are limited in scale and narrow in scope, often constrained to static, single-turn, fact-checking questions. In this work, we introduce Search Arena, a crowd-sourced, large-scale, human-preference dataset of over 24,000 paired multi-turn user interactions with search-augmented LLMs. The dataset spans diverse intents and languages, and contains full system traces with around 12,000 human preference votes. Our analysis reveals that user preferences are influenced by the number of citations, even when the cited content does not directly support the attributed claims, uncovering a gap between perceived and actual credibility. Furthermore, user preferences vary across cited sources, revealing that community-driven platforms are generally preferred and static encyclopedic sources are not always appropriate and reliable. To assess performance across different settings, we conduct cross-arena analyses by testing search-augmented LLMs in a general-purpose chat environment and conventional LLMs in search-intensive settings. We find that web search does not degrade and may even improve performance in non-search settings; however, the quality in search settings is significantly affected if solely relying on the model's parametric knowledge. We open-sourced the dataset to support future research in this direction. Our dataset and code are available at: https://github.com/lmarena/search-arena. |
Subsets and Splits