First Incorrect Step Identification on MMMU, MathVision, MathVerse-VO, DynaMath, WeMath Overall
26.4FISI F1TIM-PRM-8B
Evaluation Results
| Method | Links | |
|---|---|---|
| TIM-PRM-8BModel=TIM-PRM-8B2025.11 | 26.4 | |
| TIM-PRM-2BModel=TIM-PRM-2B2025.11 | 23.4 | |
| Qwen3-VL-8BModel=Qwen3-VL-8B2025.11 | 17.2 | |
| MM-PRM-8BModel=MM-PRM-8B2025.11 | 14.4 | |
| VisualPRM-8BModel=VisualPRM-8B2025.11 | 9.9 | |
| Qwen3-VL-2BModel=Qwen3-VL-2B2025.11 | 6 |