Scholarlink

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.

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

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

  • 会议论文

  • 发表日期：

    2026年

查看全文
  • Youssef Mohamed,
  • Mohamed Elhoseiny,
  • Thibault Formal,
  • Nadezhda Chirkova

摘要

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.