VLR-Bench: Multilingual Benchmark Dataset for Vision-Language Retrieval Augmented Generation
Lim, Hyeonseok, Shin, Dongjae, Song, Seohyun, Won, Inho, Kim, Minjun, Yuk, Junghun, Jang, Haneol, Lim, KyungTae
–arXiv.org Artificial Intelligence
We propose the VLR-Bench, a visual question answering (VQA) benchmark for evaluating vision language models (VLMs) based on retrieval augmented generation (RAG). Unlike existing evaluation datasets for external knowledge-based VQA, the proposed VLR-Bench includes five input passages. This allows testing of the ability to determine which passage is useful for answering a given query, a capability lacking in previous research. In this context, we constructed a dataset of 32,000 automatically generated instruction-following examples, which we denote as VLR-IF. This dataset is specifically designed to enhance the RAG capabilities of VLMs by enabling them to learn how to generate appropriate answers based on input passages. We evaluated the validity of the proposed benchmark and training data and verified its performance using the state-of-the-art Llama3-based VLM, the Llava-Llama-3 model. The proposed VLR-Bench and VLR-IF datasets are publicly available online.
arXiv.org Artificial Intelligence
Dec-13-2024
- Country:
- Oceania > Australia
- Queensland (0.04)
- Northern Territory (0.04)
- North America > United States
- Arizona (0.05)
- South Dakota (0.04)
- New York (0.04)
- Pennsylvania > Philadelphia County
- Philadelphia (0.04)
- Missouri > St. Louis County
- St. Louis (0.04)
- Illinois > Cook County
- Chicago (0.04)
- Hawaii > Honolulu County
- Honolulu (0.04)
- Europe
- Monaco (0.04)
- Spain > Catalonia
- Barcelona Province > Barcelona (0.14)
- Middle East > Republic of Türkiye
- Istanbul Province > Istanbul (0.04)
- Ireland > Leinster
- County Dublin > Dublin (0.04)
- Hungary > Budapest
- Budapest (0.04)
- France
- Île-de-France > Paris
- Paris (0.04)
- Bourgogne-Franche-Comté > Doubs
- Besançon (0.04)
- Île-de-France > Paris
- Asia
- China > Hong Kong (0.04)
- Singapore (0.04)
- Thailand (0.04)
- Taiwan (0.04)
- Indonesia (0.04)
- Macao (0.04)
- Malaysia (0.04)
- South Korea > Seoul
- Seoul (0.04)
- Middle East
- Jordan (0.04)
- Republic of Türkiye > Istanbul Province
- Istanbul (0.04)
- Oceania > Australia
- Genre:
- Research Report > New Finding (0.68)
- Industry:
- Government (0.93)
- Construction & Engineering (0.67)
- Leisure & Entertainment > Sports
- Baseball (1.00)
- Technology: