List your company
ResearchFormerly WizWand

Mathematical & Geometric Reasoning on We-Math

77.7Accuracy@8Qwen2.5-VL-32B + VPPO
45.14853.59962.0570.501Oct 10, 2025
Updated 8mo ago

Evaluation results

MethodLinks
Qwen2.5-VL-32B + VPPOModel Scale=32B, Algorithm=VPPO2025.1077.7
Qwen2.5-VL-32B + GRPOModel Scale=32B, Algorithm=GRPO2025.1076.7
Qwen2.5-VL-32B + DAPOModel Scale=32B, Algorithm=DAPO2025.1076.2
NoisyRollout-32BModel Scale=32B, Prompting Source=Official author-provided, Note=Trained using the training set of Geo3k2025.1075.7
MM-Eureka-32BModel Scale=32B, Prompting Source=Official author-provided2025.1075
Qwen2.5-VL-7B + VPPOModel Scale=7B, Algorithm=VPPO2025.1071.5
NoisyRollout-7BModel Scale=7B, Training Strategy=Pure RL, Prompting Source=Official author-provided, Note=Trained using the training set of Geo3k2025.1071
Qwen2.5-VL-32BModel Scale=32B2025.1071
R1-ShareVL-7BModel Scale=7B, Training Strategy=Pure RL, Prompting Source=Official author-provided2025.1069.9
VL-Rethinker-7BModel Scale=7B, Training Strategy=Pure RL, Prompting Source=Official author-provided2025.1068.9
PAPO-D-7BModel Scale=7B, Training Strategy=Pure RL2025.1068.3
Qwen2.5-VL-7B + GRPOModel Scale=7B, Algorithm=GRPO2025.1068.1
Qwen2.5-VL-7B + DAPOModel Scale=7B, Algorithm=DAPO2025.1068
ThinkLite-7BModel Scale=7B, Training Strategy=Pure RL, Prompting Source=Official author-provided2025.1066.5
MM-Eureka-7BModel Scale=7B, Training Strategy=Pure RL, Prompting Source=Official author-provided2025.1065.5
Qwen2.5-VL-7BModel Scale=7B2025.1046.4