Research map / Vision-language and multimodal / Visual question answering
M4-RAG: A Massive-Scale Multilingual Multi-Cultural Multimodal RAG
David Anugraha, Patrick Amadeus Irawan, Anshul Singh, En-Shiun Annie Lee, Genta Indra Winata
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| use multimodal retrieval (mmE5/B3) instead of caption-based text retrieval (ablation) | VLM RAG · CVQA and WorldCuisines | accuracy | Improved |
| add text-based RAG (caption query) to VLM | VLMs of various sizes · CVQA and WorldCuisines | accuracy | Hurt |
| add multimodal RAG to larger VLMs | Qwen2.5-VL 72B, Gemma3 27B · CVQA and WorldCuisines | accuracy | Mixed |
| use reasoning VLMs (Qwen3-VL) under RAG | VLM RAG · CVQA and WorldCuisines | accuracy | Improved |
| translate prompts from English to target language (ablation) | VLMs · CVQA and WorldCuisines | accuracy: -1% to -2% for high-resource languages | Hurt |
| provide oracle context in target language instead of English (ablation) | VLMs · CVQA (low-resource languages) | accuracy: -32.4% (Qwen2.5-VL 32B), -28.8% (Pangea) | Hurt |
Closest papers by meaning
- CVQA: Culturally-diverse Multilingual Visual Question Answering Benchmark (NeurIPS 2024 · similarity 0.92)
- Document Haystacks: Vision-Language Reasoning Over Piles of 1000+ Documents (CVPR 2025 · similarity 0.91)
- RAG-IGBench: Innovative Evaluation for RAG-based Interleaved Generation in Open-domain Question Answering (NeurIPS 2025 · similarity 0.90)
- IndicVisionBench: Benchmarking Cultural and Multilingual Understanding in VLMs (ICLR 2026 · similarity 0.90)
- Cross-Lingual Text-Rich Visual Comprehension: An Information Theory Perspective (AAAI 2025 · similarity 0.89)
- Multilingual Diversity Improves Vision-Language Representations (NeurIPS 2024 · similarity 0.89)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.
