List your company
ResearchFormerly WizWand

General Multimodal Understanding on MMMU Pro

56.9AccuracyQwen3-VL-32B-Instruct
14.57225.56136.5547.539May 18, 2026May 21, 2026May 24, 2026May 28, 2026May 31, 2026Jun 3, 2026Jun 7, 2026
Updated 5mo ago

Evaluation results

MethodLinks
Qwen3-VL-32B-InstructBackbone model=Qwen3-VL-32B-Instruct2026.0556.9
IVR-R1Backbone model=Qwen3-VL-4B2026.0553.3
Vision-R1Backbone model=Qwen3-VL-4B2026.0552.5
Vision-SR1Backbone model=Qwen3-VL-4B2026.0549.3
Supervised Fine-tuning (before RL)Backbone model=Qwen3-VL-4B2026.0548.6
IVR-R1Backbone model=Qwen2.5-VL-7B2026.0547.8
Vision-R1Backbone model=Qwen2.5-VL-7B2026.0547.2
Vision-SR1Backbone model=Qwen2.5-VL-7B2026.0545.5
Supervised Fine-tuning (before RL)Backbone model=Qwen2.5-VL-7B2026.0543.6
Zero-shot Inference (before RL)Backbone model=Qwen3-VL-4B2026.0542.7
VPPOModel Scale=7B, Base Model=Qwen2.5-VL-7B, Rollout=82026.0638.8
PRPOModel Scale=7B, Base Model=Qwen2.5-VL-7B, Rollout=82026.0638.8
VL-Rethinker-7BModel Scale=7B, Rollout=82026.0637.1
Qwen2.5-VL-72B-InstructBackbone model=Qwen2.5-VL-72B-Instruct2026.0536.6
PAPO-DModel Scale=7B, Base Model=Qwen2.5-VL-7B, Rollout=82026.0636.3
PAPO-GModel Scale=7B, Base Model=Qwen2.5-VL-7B, Rollout=82026.0635.5
GRPOModel Scale=7B, Base Model=Qwen2.5-VL-7B, Rollout=82026.0635.2
R1-ShareVL-7BModel Scale=7B, Rollout=82026.0635.1
NoisyRollout-7BModel Scale=7B, Rollout=82026.0634.5
PRPOModel Scale=3B, Base Model=Qwen2.5-VL-3B, Rollout=82026.0630.3
MM-Eureka-7BModel Scale=7B, Rollout=82026.0630.3
DAPOModel Scale=7B, Base Model=Qwen2.5-VL-7B, Rollout=82026.0629
PAPO-DModel Scale=3B, Base Model=Qwen2.5-VL-3B, Rollout=82026.0628.8
VPPOModel Scale=3B, Base Model=Qwen2.5-VL-3B, Rollout=82026.0628.8
DAPOModel Scale=3B, Base Model=Qwen2.5-VL-3B, Rollout=82026.0628.4
PAPO-GModel Scale=3B, Base Model=Qwen2.5-VL-3B, Rollout=82026.0626.4
GRPOModel Scale=3B, Base Model=Qwen2.5-VL-3B, Rollout=82026.0625.7
Qwen2.5-VL-7BModel Scale=7B, Rollout=82026.0625.2
Qwen2.5-VL-32B-InstructBackbone model=Qwen2.5-VL-32B-Instruct2026.0521.7
Qwen2.5-VL-3BModel Scale=3B, Rollout=82026.0619.3
Zero-shot Inference (before RL)Backbone model=Qwen2.5-VL-7B2026.0516.2