M4-RAG：大规模多语言多文化多模态检索增强生成基准
M4-RAG: A Massive-Scale Multilingual Multi-Cultural Multimodal RAG
该基准为视觉语言模型提供了跨42种语言、56种方言和189个国家的8万余个文化多样图像-问题对，并构建了可控检索环境，可用于系统评估多语言多模态RAG的检索效果与模型规模的关系，揭示大模型在RAG下性能反而下降的问题。
原文依据 We introduce M4-RAG, a massive-scale benchmark spanning 42 languages, 56 regional dialects and registers, and 189 countries
摘要译文
由模型翻译 · 术语统一 · 请对照原文核对关键结论视觉语言模型（VLM）在视觉问答（VQA）任务上已取得强劲性能，但仍受限于静态训练数据。检索增强生成（RAG）通过访问最新、具有文化根基且多语言的信息来缓解这一限制；然而，多语言多模态RAG仍很大程度上未被充分探索。我们提出了M4-RAG，一个大规模基准，涵盖42种语言、56种地区方言与语域以及189个国家，包含超过80,000个文化多样的图像-问题对，用于评估跨语言和跨模态的检索增强VQA。为了在真实性与可复现性之间取得平衡，我们构建了一个受控检索环境，其中包含数百万份精心整理的、与查询领域相关的多语言文档，在确保实验一致性的同时近似真实世界的检索条件。我们的系统性评估揭示，尽管RAG能持续惠及较小的VLM，但它无法扩展至更大的模型，甚至常常降低其性能，暴露出模型规模与当前检索有效性之间的关键失配。我们的跨语言评估还揭示，当提示或检索上下文以非英语语言提供时，性能会出现显著下降。M4-RAG的代码、数据集和评估协议已在https://github.com/davidanugraha/M4-RAG开源。
英文原文摘要 1427 字符
Vision-language models (VLMs) have achieved strong performance in visual question answering (VQA), yet they remain constrained by static training data. Retrieval-Augmented Generation (RAG) mitigates this limitation by enabling access to up-to-date, culturally grounded, and multilingual information; however, multilingual multimodal RAG remains largely underexplored. We introduce M4-RAG, a massive-scale benchmark spanning 42 languages, 56 regional dialects and registers, and 189 countries, comprising over 80,000 culturally diverse image-question pairs for evaluating retrieval-augmented VQA across languages and modalities. To balance realism with reproducibility, we build a controlled retrieval environment containing millions of carefully curated multilingual documents relevant to the query domains, approximating real-world retrieval conditions while ensuring consistent experimentation. Our systematic evaluation reveals that although RAG consistently benefits smaller VLMs, it fails to scale to larger models and often even degrades their performance, exposing a critical mismatch between model size and current retrieval effectiveness. Our cross-lingual evaluations also reveal significant performance degradation when prompts or retrieved context are provided in non-English languages. The code, datasets, and evaluation protocols for M4-RAG are available as open-source at https://github.com/davidanugraha/M4-RAG.