General Multimodal Understanding on MMMU Pro

56.9Accuracy

Qwen3-VL-32B-Instruct

Updated 1mo ago

Evaluation Results

Method	Links
Qwen3-VL-32B-Instruct 2026.05		56.9
IVR-R1 2026.05		53.3
Vision-R1 2026.05		52.5
Vision-SR1 2026.05		49.3
Supervised Fine-tuning (before RL) 2026.05		48.6
IVR-R1 2026.05		47.8
Vision-R1 2026.05		47.2
Vision-SR1 2026.05		45.5
Supervised Fine-tuning (before RL) 2026.05		43.6
Zero-shot Inference (before RL) 2026.05		42.7
VPPO 2026.06		38.8
PRPO 2026.06		38.8
VL-Rethinker-7B 2026.06		37.1
Qwen2.5-VL-72B-Instruct 2026.05		36.6
PAPO-D 2026.06		36.3
PAPO-G 2026.06		35.5
GRPO 2026.06		35.2
R1-ShareVL-7B 2026.06		35.1
NoisyRollout-7B 2026.06		34.5
PRPO 2026.06		30.3
MM-Eureka-7B 2026.06		30.3
DAPO 2026.06		29
PAPO-D 2026.06		28.8
VPPO 2026.06		28.8
DAPO 2026.06		28.4
PAPO-G 2026.06		26.4
GRPO 2026.06		25.7
Qwen2.5-VL-7B 2026.06		25.2
Qwen2.5-VL-32B-Instruct 2026.05		21.7
Qwen2.5-VL-3B 2026.06		19.3
Zero-shot Inference (before RL) 2026.05		16.2