CroSearch-R1: Better Leveraging Cross-lingual Knowledge for Retrieval-Augmented Generation

TL;DR

CroSearch-R1 is a novel framework that enhances multilingual retrieval-augmented generation by dynamically integrating cross-lingual knowledge through reinforcement learning and multi-turn retrieval strategies.

Contribution

It introduces a search-augmented reinforcement learning approach with multilingual retrieval and a rollout mechanism to better leverage cross-lingual knowledge in RAG.

Findings

Improves RAG effectiveness with multilingual collections.

Effectively leverages cross-lingual complementarity.

Demonstrates superior performance in experiments.

Abstract

A multilingual collection may contain useful knowledge in other languages to supplement and correct the facts in the original language for Retrieval-Augmented Generation (RAG). However, the vanilla approach that simply concatenates multiple pieces of knowledge from different languages into the context may fail to improve effectiveness due to the potential disparities across languages. To better leverage multilingual knowledge, we propose CroSearch-R1, a search-augmented reinforcement learning framework to integrate multilingual knowledge into the Group Relative Policy Optimization (GRPO) process. In particular, the approach adopts a multi-turn retrieval strategy with cross-lingual knowledge integration to dynamically align the knowledge from other languages as supplementary evidence into a unified representation space. Furthermore, we introduce a multilingual rollout mechanism to optimize reasoning transferability across languages. Experimental results demonstrate that our framework effectively leverages cross-lingual complementarity and improves the effectiveness of RAG with multilingual collections.