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

2026-01-26

AI Summary

This work addresses the inefficiency of multilingual retrieval-augmented generation (RAG) systems caused by redundant context, a challenge exacerbated by the limited cross-lingual generalization of existing pruning methods. To this end, we propose XProvence, the first zero-cost multilingual context pruning approach that seamlessly integrates pruning capability directly into the reranker without incurring additional computational overhead. Building upon the Provence framework, XProvence leverages multilingual pretraining and cross-lingual transfer to support over 100 languages. Extensive experiments on four multilingual question answering benchmarks demonstrate that XProvence achieves substantial context compression with negligible performance degradation, significantly outperforming strong baselines.

Technology Category

• Natural Language Processing: Machine Translation, Multilinguality, Cross-Lingual NLP
• Machine Learning: Large Multimodal Models (LMMs)
• Data Mining & Knowledge Management: Conversational Systems for Recommendation & Retrieval

Application Category

• Search and Retrieval-Augmented AI: Multilingual and cross-lingual Web search
• Web Mining and Content Analysis: Mining multimedia, multimodal, multilingual, cross-lingual Web data
• User Modeling, Personalization and Recommendation: Fairness-aware retrieval and ranking

Abstract

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.

Problem

• multilingual
• context pruning
• retrieval-augmented generation
• zero-cost
• RAG

Innovation

• zero-cost pruning
• multilingual RAG
• cross-lingual transfer
• context pruning
• retrieval-augmented generation