CoolFace
Datasetpublic

uclanlp/MRAG-Bench

MRAG-Bench: Vision-Centric Evaluation for Retrieval-Augmented Multimodal Models ๐ŸŒ Homepage | ๐Ÿ“– Paper | ๐Ÿ’ป Evaluation Intro MRAG-Bench consists of 16,130 images and 1,353 human-annotated multiple-choice questions across 9 distinct scenarios, providing a robust and systematic evaluation of Large Vision Language Model (LVLM)โ€™s vision-centric multimodal retrieval-augmented generation (RAG) abilities. Results Evaluated upon 10 open-source and 4โ€ฆ See the full description on the dataset page: https://huggingface.co/datasets/uclanlp/MRAG-Bench.

sourceHugging Facecc-by-4.0updated 2y agoView on Hugging Face
13likes605downloads

uclanlp/MRAG-Bench ยท main ยท files are served by the source, never re-hosted here