VRBench

What is VRBench?

VRBench is the first evaluation benchmark specifically designed to assess multi-step reasoning capabilities over long videos. The benchmark comprises 1,010…

Released

2025-06-12

Modality

multimodal

Evaluates

多模态, Multimodal, 推理, Reasoning, 视频理解, Video-Understanding, 多步推理, 视频推理, 多语种, 多模态模型, VLM, 逻辑推理, Video Understanding, 不支持, Unsupported

Openness

open

Importer

OpenCompass Hub

Reported scores

0

Source provenance

• Original evidence https://hub.opencompass.org.cn/dataset-detail/VRBench

VRBench paper, code and dataset

• Paper https://arxiv.org/abs/2506.10857
• Code https://github.com/OpenGVLab/VRBench
• Dataset https://huggingface.co/datasets/OpenGVLab/VRBench

Which sources cite VRBench?

• VRBench

Related 多模态, Multimodal, 推理, Reasoning, 视频理解, Video-Understanding, 多步推理, 视频推理, 多语种, 多模态模型, VLM, 逻辑推理, Video Understanding, 不支持, Unsupported benchmarks

• 3DCodeBench
• A-Bench
• A-OKVQA
• AbsPyramid
• ActionAtlas
• ActiveFly-Bench
• Ada-LEval
• AdvancedMathBench