Research map / Vision-language and multimodal / Visual question answering
RAVENEA: A Benchmark for Multimodal Retrieval-Augmented Visual Culture Understanding
Jiaang Li, Yifei Yuan, Wenyan Li, Mohammad Aliannejadi, Daniel Hershcovich, Anders Søgaard, Ivan Vulić, Wenxuan Zhang, Paul Pu Liang, Yang Deng, Serge Belongie
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 |
|---|---|---|---|
| fine-tune CLIP with culture-aware contrastive loss on RAVENEA | CLIP retriever · RAVENEA retrieval | P@1: 60.87 -> 72.05 | Improved |
| fine-tune SigLIP with culture-aware contrastive loss | SigLIP retriever · RAVENEA retrieval | MRR: 68.62 -> 70.95 | Improved |
| fine-tune generative models (VisualBERT, VL-T5, LLaVA-OV-7B) as retrievers instead of contrastive models | retriever · RAVENEA retrieval | MRR: 75.44 -> 61.33 | Hurt |
| add Ravenea-CLIP retrieval augmentation to VLMs | 17 VLMs · cVQA | accuracy: 67.1 -> 73.3 | Improved |
| add Ravenea-CLIP retrieval augmentation to VLMs | 17 VLMs · cIC | RegionScore: 57.5 -> 68.8 | Improved |
| use culture-aware fine-tuned CLIP instead of frozen CLIP for retrieval | VLMs with RAG · cVQA | accuracy: 71.9 -> 73.3 | Improved |
| use culture-aware fine-tuned CLIP instead of frozen CLIP for retrieval | VLMs with RAG · cIC | RegionScore: 65.1 -> 68.8 | Improved |
| add Ravenea-CLIP retrieval to a small VLM | Qwen3-VL-2B · cVQA | accuracy: +31.6% | Improved |
| add Ravenea-CLIP retrieval to a large VLM | Qwen3-VL-32B · cIC | RegionScore: no gain | No clear effect |
| add culturally relevant retrieved documents instead of unrelated documents (ablation) | Ravenea-CLIP · cVQA (12 VLMs) | accuracy: 60.6 -> 71.2 | Improved |
| add culturally relevant retrieved documents instead of random characters (ablation) | Ravenea-CLIP · cVQA (12 VLMs) | accuracy: 62.3 -> 71.2 | Improved |
| add culturally relevant retrieved documents instead of unrelated documents (ablation) | Ravenea-CLIP · cIC (12 VLMs) | RegionScore: 48.8 -> 67.7 | Improved |
Closest papers by meaning
- Culture In a Frame: C$^3$B as a Comic-Based Benchmark for Multimodal Culturally Awareness (ICLR 2026 · similarity 0.90)
- CVQA: Culturally-diverse Multilingual Visual Question Answering Benchmark (NeurIPS 2024 · similarity 0.90)
- IndicVisionBench: Benchmarking Cultural and Multilingual Understanding in VLMs (ICLR 2026 · similarity 0.89)
- Beyond Aesthetics: Cultural Competence in Text-to-Image Models (NeurIPS 2024 · similarity 0.88)
- World in a Frame: Understanding Culture Mixing as a New Challenge for Vision-Language Models (CVPR 2026 · similarity 0.88)
- See It from My Perspective: How Language Affects Cultural Bias in Image Understanding (ICLR 2025 · similarity 0.88)
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.
