List your company
ResearchFormerly WizWand

Mathematical & Geometric Reasoning on DynaMath (accuracy@8)

73.1Accuracy@8Qwen2.5-VL-32B + VPPO
55.00459.70264.469.098Oct 10, 2025
Updated 8mo ago

Evaluation results

MethodLinks
Qwen2.5-VL-32B + VPPOModel Scale=32B, Algorithm=VPPO2025.1073.1
Qwen2.5-VL-32B + DAPOModel Scale=32B, Algorithm=DAPO2025.1072.6
NoisyRollout-32BModel Scale=32B, Prompting Source=Official author-provided, Note=Trained using the training set of Geo3k2025.1072.2
MM-Eureka-32BModel Scale=32B, Prompting Source=Official author-provided2025.1072
Qwen2.5-VL-32B + GRPOModel Scale=32B, Algorithm=GRPO2025.1071.6
Qwen2.5-VL-32BModel Scale=32B2025.1068.7
Qwen2.5-VL-7B + VPPOModel Scale=7B, Algorithm=VPPO2025.1068.1
PAPO-D-7BModel Scale=7B, Training Strategy=Pure RL2025.1066.8
Qwen2.5-VL-7B + DAPOModel Scale=7B, Algorithm=DAPO2025.1066.6
Qwen2.5-VL-7B + GRPOModel Scale=7B, Algorithm=GRPO2025.1065.8
VL-Rethinker-7BModel Scale=7B, Training Strategy=Pure RL, Prompting Source=Official author-provided2025.1065.7
NoisyRollout-7BModel Scale=7B, Training Strategy=Pure RL, Prompting Source=Official author-provided, Note=Trained using the training set of Geo3k2025.1065.5
MM-Eureka-7BModel Scale=7B, Training Strategy=Pure RL, Prompting Source=Official author-provided2025.1065.4
R1-ShareVL-7BModel Scale=7B, Training Strategy=Pure RL, Prompting Source=Official author-provided2025.1065.1
ThinkLite-7BModel Scale=7B, Training Strategy=Pure RL, Prompting Source=Official author-provided2025.1064.6
Qwen2.5-VL-7BModel Scale=7B2025.1055.7