Multimodal Reasoning on MMMU (Strict Accuracy, Standard Accuracy, False Positive) benchmark leaderboard · SOTA2 Research

Multimodal Reasoning on MMMU (Strict Accuracy, Standard Accuracy, False Positive)

54.67Strict AccuracyAutoRubric

Evaluation Results

Method	Links
AutoRubric2025.10		54.67	56.56	-1.89
NoisyRollout2025.10		53.22	56.11	-2.89
VL-Rethinker2025.10		52.22	57.11	-4.89
VL-Reasoner2025.10		52	56.22	-4.22
MM-Eureka2025.10		51	52.78	-1.78
Perception-R12025.10		50.89	52.89	-2
ThinkLite-VL2025.10		50.56	53.67	-3.11
Qwen2.5-VL-7B2025.10		49.8	55	-5.2
VLAA-Thinker2025.10		48	52.33	-4.33
Vision-G12025.10		47.67	53.78	-6.11
R1-VL2025.10		42	46.56	-4.56