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

Spatial Reasoning (Multi-Image) on SPAR-Bench

67.3Accuracy

Human

Updated 7d ago

Evaluation Results

Method	Links
Human 2025.10		67.3
ViLAVT-7B 2026.02		52.6
SpaceVista-7B 2025.10		41.6
AdaTooler-V-7B 2025.12		40.3
SpaceVista-7B 2025.10		38.1
VILASR-7B 2026.02		37.6
VILASR 2025.12		37.6
SpaceR-7B 2025.10		37.6
VILASR-7B 2025.10		37.6
GPT-5 2025.10		37.4
SpaceR-7B 2026.02		37.1
Qwen2.5-VL-7B 2026.02		36.9
Qwen2.5-VL-7B 2025.10		36.9
Gemini-2.5-pro 2025.10		36.3
InternVL3-8B 2026.02		36
InternVL3.5-8B 2025.10		36
Spatial-MLLM-4B 2026.02		35.1
GPT-4o 2025.12		33.6
Qwen2.5-VL-7B 2025.10		33.1
Qwen2.5-VL-7B-Instruct 2025.12		33.07
LLaVA-OneVision-7B 2026.02		32.4
Qwen2.5-VL-72B 2025.10		32.4
Qwen2.5-VL-7B 2026.02		31.6
SpatialMLLM-4B 2025.10		31.5
LLaVA-NeXT-Video-7B 2025.10		31.3
InternVL3.5-38B 2025.10		31
LLAVA-Onevision-7B 2025.10		30.6
LLaVA-1.5-7B 2025.12		23.65