XProvence: Zero-Cost Multilingual Context Pruning for Retrieval-Augmented Generation

Fuente: arXiv
Saved in:
Bibliographic Details
Main Authors: Mohamed, Youssef, Elhoseiny, Mohamed, Formal, Thibault, Chirkova, Nadezhda
Format: Preprint
Published: 2026
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866910001715478528
author Mohamed, Youssef
Elhoseiny, Mohamed
Formal, Thibault
Chirkova, Nadezhda
author_facet Mohamed, Youssef
Elhoseiny, Mohamed
Formal, Thibault
Chirkova, Nadezhda
contents This paper introduces XProvence, a multilingual zero-cost context pruning model for retrieval-augmented generation (RAG), trained on 16 languages and supporting 100+ languages through effective cross-lingual transfer. Motivated by the growing use of RAG systems across diverse languages, we explore several strategies to generalize the Provence framework-which first integrated efficient zero-cost context pruning directly into the re-ranking model-beyond English. Across four multilingual question answering benchmarks, we show how XProvence can prune RAG contexts with minimal-to-no performance degradation and outperforms strong baselines. Our model is available at https://huggingface.co/naver/xprovence-reranker-bgem3-v2.
format Preprint
id arxiv_https___arxiv_org_abs_2601_18886
institution arXiv
publishDate 2026
record_format arxiv
spellingShingle XProvence: Zero-Cost Multilingual Context Pruning for Retrieval-Augmented Generation
Mohamed, Youssef
Elhoseiny, Mohamed
Formal, Thibault
Chirkova, Nadezhda
Information Retrieval
Computation and Language
This paper introduces XProvence, a multilingual zero-cost context pruning model for retrieval-augmented generation (RAG), trained on 16 languages and supporting 100+ languages through effective cross-lingual transfer. Motivated by the growing use of RAG systems across diverse languages, we explore several strategies to generalize the Provence framework-which first integrated efficient zero-cost context pruning directly into the re-ranking model-beyond English. Across four multilingual question answering benchmarks, we show how XProvence can prune RAG contexts with minimal-to-no performance degradation and outperforms strong baselines. Our model is available at https://huggingface.co/naver/xprovence-reranker-bgem3-v2.
title XProvence: Zero-Cost Multilingual Context Pruning for Retrieval-Augmented Generation
topic Information Retrieval
Computation and Language
url https://arxiv.org/abs/2601.18886