Share your thoughts, 1 month free Claude Pro on usSee more

Video Spatial Reasoning on STI-Bench

41.4Accuracy

Gemini-2.5-pro

Updated 1mo ago

Evaluation Results

Method	Links
Gemini-2.5-pro 2025.10		41.4
Qwen2.5-VL-72B 2025.10		40.7
GPT-5 2025.10		39.3
InternVL3.5-38B 2025.10		39.2
SpaceVista-7B 2025.10		38.2
SpaceR-7B 2025.10		37
SpaceVista-7B 2025.10		35.9
Qwen2.5-VL-7B 2025.10		35
InternVL3.5-8B 2025.10		33.2
Qwen2.5-VL-7B 2025.10		32.1
VILASR-7B 2025.10		31.5
SpatialMLLM-4B 2025.10		30.5
LLaVA-NeXT-Video-7B 2025.10		29.9
VG LLM-4B 2025.10		29.3
LLAVA-Onevision-7B 2025.10		29