naver
/

xprovence-reranker-bgem3-v1

@@ -30,12 +30,14 @@ pipeline_tag: text-ranking
 # Model Card for XProvence-reranker
 XProvence is a Zero Cost **context pruning model** that seamlessly integrates with reranker for retrieval-augmented generation,
 particularly **optimized for question answering**. Given a user question and a retrieved passage, XProvence **removes sentences
 from the passage that are not relevant to the user question**. This **speeds up generation** and **reduces context noise**, in
 a plug-and-play manner **for any LLM**.
-XProvence extends Provence by supporting 16 languages natively. It also supports 100+ languages through cross lingual transfer, since
 it is based on BGE-m3 which is pretrained on 100+ languages.
@@ -111,8 +113,8 @@ Interface of the `process` function:
 * Input: user question (e.g., a sentence) + retrieved context passage (e.g., a paragraph)
 * Output: pruned context passage, i.e., irrelevant sentences are removed + relevance score (can be used for reranking)
 * Model Architecture: The model was initialized from [bge-reranker-v2-m3](https://huggingface.co/BAAI/bge-reranker-v2-m3) and finetuned with two objectives: (1) output a binary mask which can be used to prune irrelevant sentences; and (2) preserve initial reranking capabilities.
-* Training data: Training queries from [MIRACL](https://huggingface.co/datasets/miracl/miracl) and documents from [Wikipedia](https://huggingface.co/datasets/wikimedia/wikipedia), with synthetic silver labelling of which sentences to keep, produced using [aya-expanse-8b](https://huggingface.co/CohereLabs/aya-expanse-8b).
-* Languages covered: Arabic, Bengali, English, Spanish, Persian, Finnish, France, Hindi, Indonesian, Japanese, Korean, Russian, Swahili, Telugu, Thai, Chinese
 * Context length: 8192 tokens (similar to the pretrained BGE-m3 model)
 * Evaluation: we evaluate XProvence on 26 languages from 6 different datasets. We find that XProvence is able to prune irrelevant sentences with little-to-no drop in performance, on all languages, and outperforms existing baselines on the Pareto front.

 # Model Card for XProvence-reranker
+<img src="https://cdn-uploads.huggingface.co/production/uploads/6273df31c3b822dad2d1eef2/4n-bxYfiMPC2LoLM2m7pg.png" alt="image/png" width="600">
 XProvence is a Zero Cost **context pruning model** that seamlessly integrates with reranker for retrieval-augmented generation,
 particularly **optimized for question answering**. Given a user question and a retrieved passage, XProvence **removes sentences
 from the passage that are not relevant to the user question**. This **speeds up generation** and **reduces context noise**, in
 a plug-and-play manner **for any LLM**.
+XProvence is a multilingual version of [Provence](https://huggingface.co/naver/provence-reranker-debertav3-v1) supporting 16 languages natively. It also supports 100+ languages through cross lingual transfer, since
 it is based on BGE-m3 which is pretrained on 100+ languages.
 * Input: user question (e.g., a sentence) + retrieved context passage (e.g., a paragraph)
 * Output: pruned context passage, i.e., irrelevant sentences are removed + relevance score (can be used for reranking)
 * Model Architecture: The model was initialized from [bge-reranker-v2-m3](https://huggingface.co/BAAI/bge-reranker-v2-m3) and finetuned with two objectives: (1) output a binary mask which can be used to prune irrelevant sentences; and (2) preserve initial reranking capabilities.
+* Training data: [MIRACL](https://huggingface.co/datasets/miracl/miracl), with synthetic silver labelling of which sentences to keep, produced using [aya-expanse-8b](https://huggingface.co/CohereLabs/aya-expanse-8b).
+* Languages in the training data: Arabic, Bengali, English, Spanish, Persian, Finnish, France, Hindi, Indonesian, Japanese, Korean, Russian, Swahili, Telugu, Thai, Chinese
 * Context length: 8192 tokens (similar to the pretrained BGE-m3 model)
 * Evaluation: we evaluate XProvence on 26 languages from 6 different datasets. We find that XProvence is able to prune irrelevant sentences with little-to-no drop in performance, on all languages, and outperforms existing baselines on the Pareto front.