Step-wise Verification on WeMath
63.9Macro F1TIM-PRM
Evaluation Results
| Method | Links | |
|---|---|---|
| TIM-PRMModel Type=Open-source, Number of Parameters=8B2025.11 | 63.9 | |
| GPT-4oModel Type=Proprietary2025.11 | 63.3 | |
| Qwen3-VLModel Type=Open-source, Number of Parameters=8B2025.11 | 62.7 | |
| Qwen2.5-VLModel Type=Open-source, Number of Parameters=72B2025.11 | 62.3 | |
| Gemini-2.0-FlashModel Type=Proprietary2025.11 | 58.7 | |
| GPT-4o-MiniModel Type=Proprietary2025.11 | 58.5 | |
| Qwen3-VLModel Type=Open-source, Number of Parameters=2B2025.11 | 58.3 | |
| TIM-PRMModel Type=Open-source, Number of Parameters=2B2025.11 | 58 | |
| MiniCPM-V2.6Model Type=Open-source, Number of Parameters=8B2025.11 | 57.4 | |
| LLaVA-OVModel Type=Open-source, Number of Parameters=72B2025.11 | 57.3 | |
| MM-PRMModel Type=Open-source, Number of Parameters=8B2025.11 | 56.5 | |
| VisualPRMModel Type=Open-source, Number of Parameters=8B2025.11 | 55.1 | |
| Qwen2.5-VLModel Type=Open-source, Number of Parameters=7B2025.11 | 54.2 | |
| LLaVA-OVModel Type=Open-source, Number of Parameters=7B2025.11 | 52.5 | |
| InternVL2.5Model Type=Open-source, Number of Parameters=38B2025.11 | 52.5 | |
| InternVL2.5Model Type=Open-source, Number of Parameters=78B2025.11 | 52.5 | |
| InternVL2.5Model Type=Open-source, Number of Parameters=26B2025.11 | 51.4 | |
| InternVL2.5Model Type=Open-source, Number of Parameters=8B2025.11 | 50.8 |