Math Reasoning on GSM8K (Calibration Metrics)
0.87AUROCVEROIC
Evaluation Results
| Method | Links | |||||
|---|---|---|---|---|---|---|
| VEROICsetting=CMIS, inference_path=LLaMA-3.1-8B -> LLaMA-3.1-70B2026.04 | 0.87 | 0.52 | 0.139 | 0.417 | 0.048 | |
| RISK PREDICTORsetting=CMIS, inference_path=LLaMA-3.1-8B -> LLaMA-3.1-70B2026.04 | 0.85 | 0.51 | 0.144 | 0.432 | 0.058 | |
| CONFIDENCE-THRESHOLDsetting=CMIS, inference_path=LLaMA-3.1-8B -> LLaMA-3.1-70B2026.04 | 0.84 | 0.49 | 0.152 | 0.452 | 0.066 | |
| OUTPUT-TRIGGERsetting=CMIS, inference_path=LLaMA-3.1-8B -> LLaMA-3.1-70B2026.04 | 0.83 | 0.48 | 0.149 | 0.447 | 0.062 | |
| PERIODICsetting=CMIS, inference_path=LLaMA-3.1-8B -> LLaMA-3.1-70B2026.04 | 0.77 | 0.43 | 0.162 | 0.467 | 0.065 | |
| RANDOMsetting=CMIS, inference_path=LLaMA-3.1-8B -> LLaMA-3.1-70B2026.04 | 0.73 | 0.41 | 0.173 | 0.475 | 0.076 | |
| FIXEDsetting=CMIS, inference_path=LLaMA-3.1-8B -> LLaMA-3.1-70B2026.04 | 0.71 | 0.32 | 0.185 | 0.523 | 0.085 |