List your company

Video Question Answering on UrbanVideo and VSI-Bench Clean

62Original MetricGemini-1.5-Pro
32.8840.444855.56Mar 11, 2026

Evaluation results

MethodLinks
Gemini-1.5-ProSize=–, #Fr=322026.0362
GPT-4oSize=–, #Fr=322026.0359
ROVASize=72B, #Fr=32, Backbone=Qwen2.5-VL2026.0359
Video-R1Size=72B, #Fr=322026.0358
Qwen2.5-VLSize=72B, #Fr=322026.0357
ROVASize=13B, #Fr=32, Backbone=Qwen2-VL2026.0356
ROVASize=7B, #Fr=32, Backbone=Embodied-R2026.0355
Embodied-RSize=7B, #Fr=322026.0354
Claude-3.5-Son.Size=–, #Fr=322026.0353
ROVASize=7B, #Fr=32, Backbone=Qwen2.5-VL2026.0353
Qwen2-VLSize=13B, #Fr=322026.0353
Video-R1Size=7B, #Fr=322026.0351
Qwen2.5-VLSize=7B, #Fr=322026.0351
LLaVA-Video-RSize=7B, #Fr=322026.0348
ROVASize=8B, #Fr=32, Backbone=InternVL2.52026.0347
InternVL2.5Size=8B, #Fr=322026.0346
VideoChat-RSize=7B, #Fr=162026.0345
LLaVA-VideoSize=7B, #Fr=322026.0344
MiniCPM-V 2.6Size=8B, #Fr=642026.0343
VideoLLaMA2Size=7B, #Fr=162026.0336
VideoChat2Size=7B, #Fr=162026.0334