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

Spatial Reasoning on CVBench

96.312D Relationship Score

RoboRefer-8B-SFT

Updated 1mo ago

Evaluation Results

Method	Links
RoboRefer-8B-SFT 2025.12		96.31	96.5	-
GPT-5 2025.12		95.54	91.33	-
Gemini-ER 1.5 2025.12		95.54	90.5	-
SpaceTools-3B 2025.12		94.92	96	-
Qwen2.5-VL-3B-Tool-free 2025.12		91.54	83.33	-
Qwen2.5-VL-32B 2025.12		90.46	86.67	-
Claude Sonnet 4.5 2025.12		89.85	78.5	-
GPT-4o 2025.12		88.77	75.5	-
Qwen2.5-VL-3B-Tool-free 2025.12		87.38	70.83	-
RoboBrain2.0-7B 2025.12		87.23	90	-
RoboPoint-13B 2025.12		74	76.5	-
LLaVA-NeXT-8B 2025.12		72.15	73.67	-
Molmo-7B 2025.12		72.15	73.33	-
Qwen2.5-VL-3B 2025.12		70.62	65.33	-
SpaceLLaVA-13B 2025.12		61.08	62.83	-
GPT-5.2 2026.03		-	-	79.62
Gemini-3-flash 2026.03		-	-	83.94
Embodied-R1-3B 2026.03		-	-	81.69
PEEK-3B 2026.03		-	-	55.49
Robobrain2.0-3B 2026.03		-	-	81.22
Qwen3v1-4B 2026.03		-	-	79.21
ST-VLA 2026.03		-	-	84.52