Mathematical & Geometric Reasoning on Geo3k
56.6Accuracy@8NoisyRollout-32B
Evaluation Results
| Method | Links | |
|---|---|---|
| NoisyRollout-32BModel Scale=32B, Prompting Source=Official author-provided, Note=Trained using the training set of Geo3k2025.10 | 56.6 | |
| Qwen2.5-VL-32B + VPPOModel Scale=32B, Algorithm=VPPO2025.10 | 53.4 | |
| NoisyRollout-7BModel Scale=7B, Training Strategy=Pure RL, Prompting Source=Official author-provided, Note=Trained using the training set of Geo3k2025.10 | 51.8 | |
| Qwen2.5-VL-32B + GRPOModel Scale=32B, Algorithm=GRPO2025.10 | 51.4 | |
| Qwen2.5-VL-32B + DAPOModel Scale=32B, Algorithm=DAPO2025.10 | 51.4 | |
| MM-Eureka-32BModel Scale=32B, Prompting Source=Official author-provided2025.10 | 51 | |
| Qwen2.5-VL-32BModel Scale=32B2025.10 | 47 | |
| Qwen2.5-VL-7B + VPPOModel Scale=7B, Algorithm=VPPO2025.10 | 46.5 | |
| PAPO-D-7BModel Scale=7B, Training Strategy=Pure RL2025.10 | 44.1 | |
| Qwen2.5-VL-7B + DAPOModel Scale=7B, Algorithm=DAPO2025.10 | 41.5 | |
| R1-ShareVL-7BModel Scale=7B, Training Strategy=Pure RL, Prompting Source=Official author-provided2025.10 | 41.2 | |
| VL-Rethinker-7BModel Scale=7B, Training Strategy=Pure RL, Prompting Source=Official author-provided2025.10 | 40.7 | |
| MM-Eureka-7BModel Scale=7B, Training Strategy=Pure RL, Prompting Source=Official author-provided2025.10 | 40.3 | |
| Qwen2.5-VL-7B + GRPOModel Scale=7B, Algorithm=GRPO2025.10 | 40.2 | |
| ThinkLite-7BModel Scale=7B, Training Strategy=Pure RL, Prompting Source=Official author-provided2025.10 | 37.6 | |
| Qwen2.5-VL-7BModel Scale=7B2025.10 | 37.1 |