一作ACL 2026Poster· Natural Language Processing
All Languages Matter: Understanding and Mitigating Language Bias in Multilingual RAG
Dan Wang · Guozhao Mo · Yafei Shi · Cheng Zhang · Bo Zheng (Taobao & · Boxi Cao · Xuanang Chen · Yaojie Lu · Hongyu Lin · Ben He · Xianpei Han · Le Sun
摘要
Multilingual Retrieval-Augmented Generation (mRAG) leverages cross-lingual evidence to ground Large Language Models (LLMs) in global knowledge. However, we show that current mRAG systems suffer from a language bias during reranking, systematically favoring English and the query’s native language....