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

Multi-discipline Reasoning on MMMU-Pro (accuracy@8)

47.1Accuracy@8

Qwen2.5-VL-32B + VPPO

Updated 3mo ago

Evaluation Results

Method	Links
Qwen2.5-VL-32B + VPPO 2025.10		47.1
Qwen2.5-VL-32B + DAPO 2025.10		46.4
Qwen2.5-VL-32B + GRPO 2025.10		45.4
MM-Eureka-32B 2025.10		43.1
NoisyRollout-32B 2025.10		43.1
Qwen2.5-VL-32B 2025.10		39.6
Qwen2.5-VL-7B + VPPO 2025.10		37.9
VL-Rethinker-7B 2025.10		37
PAPO-D-7B 2025.10		36.3
Qwen2.5-VL-7B + DAPO 2025.10		35.9
Qwen2.5-VL-7B + GRPO 2025.10		35.2
R1-ShareVL-7B 2025.10		35.1
NoisyRollout-7B 2025.10		34.5
MM-Eureka-7B 2025.10		30.3
ThinkLite-7B 2025.10		28
Qwen2.5-VL-7B 2025.10		25.1