List your company
ResearchFormerly WizWand

Mathematical Reasoning on MMK12

82.8AccuracyPRPO
35.27247.61159.9572.289Jun 7, 2026
Updated 5mo ago

Evaluation results

MethodLinks
PRPOModel Scale=7B, Base Model=Qwen2.5-VL-7B, Rollout=82026.0682.8
PAPO-DModel Scale=7B, Base Model=Qwen2.5-VL-7B, Rollout=82026.0680.6
VPPOModel Scale=7B, Base Model=Qwen2.5-VL-7B, Rollout=82026.0680.1
DAPOModel Scale=7B, Base Model=Qwen2.5-VL-7B, Rollout=82026.0675.9
GRPOModel Scale=7B, Base Model=Qwen2.5-VL-7B, Rollout=82026.0672.3
PAPO-GModel Scale=7B, Base Model=Qwen2.5-VL-7B, Rollout=82026.0671.8
R1-ShareVL-7BModel Scale=7B, Rollout=82026.0670.9
VL-Rethinker-7BModel Scale=7B, Rollout=82026.0668.3
MM-Eureka-7BModel Scale=7B, Rollout=82026.0667.7
PRPOModel Scale=3B, Base Model=Qwen2.5-VL-3B, Rollout=82026.0666.9
DAPOModel Scale=3B, Base Model=Qwen2.5-VL-3B, Rollout=82026.0666.8
VPPOModel Scale=3B, Base Model=Qwen2.5-VL-3B, Rollout=82026.0665.3
PAPO-DModel Scale=3B, Base Model=Qwen2.5-VL-3B, Rollout=82026.0664.1
GRPOModel Scale=3B, Base Model=Qwen2.5-VL-3B, Rollout=82026.0657.2
PAPO-GModel Scale=3B, Base Model=Qwen2.5-VL-3B, Rollout=82026.0656.3
NoisyRollout-7BModel Scale=7B, Rollout=82026.0651
Qwen2.5-VL-7BModel Scale=7B, Rollout=82026.0643.3
Qwen2.5-VL-3BModel Scale=3B, Rollout=82026.0637.1