Mathematical Reasoning on MATH-500 (Accuracy and Calibration Metrics)
77AccuracyVanilla SimpleTIR
Evaluation Results
| Method | Links | ||||
|---|---|---|---|---|---|
| Vanilla SimpleTIRModel=Qwen2.5-3B2026.01 | 77 | 0.151 | 0.193 | 0.622 | |
| CAR (MSCR)Model=Qwen2.5-3B2026.01 | 76.9 | 0.057 | 0.168 | 0.636 |
| Method | Links | ||||
|---|---|---|---|---|---|
| Vanilla SimpleTIRModel=Qwen2.5-3B2026.01 | 77 | 0.151 | 0.193 | 0.622 | |
| CAR (MSCR)Model=Qwen2.5-3B2026.01 | 76.9 | 0.057 | 0.168 | 0.636 |