Uncertainty Calibration on GPQA main
0.672AUROCPE
Evaluation Results
| Method | Links | |
|---|---|---|
| PEBackbone=Qwen2.5-32B-Instruct2026.01 | 0.672 | |
| p(True)Backbone=Qwen2.5-7B-Instruct2026.01 | 0.649 | |
| ET-PEBackbone=Qwen2.5-32B-Instruct2026.01 | 0.636 | |
| PEBackbone=Qwen2.5-14B-Instruct2026.01 | 0.635 | |
| LSBackbone=Qwen2.5-32B-Instruct2026.01 | 0.634 | |
| ET-PEBackbone=Qwen2.5-14B-Instruct2026.01 | 0.622 | |
| LSBackbone=Qwen2.5-14B-Instruct2026.01 | 0.613 | |
| LSBackbone=Qwen2.5-7B-Instruct2026.01 | 0.601 | |
| p(True)Backbone=Qwen2.5-32B-Instruct2026.01 | 0.591 | |
| ET-PEBackbone=Qwen2.5-7B-Instruct2026.01 | 0.59 | |
| PEBackbone=Qwen2.5-7B-Instruct2026.01 | 0.586 | |
| p(True)Backbone=Qwen2.5-14B-Instruct2026.01 | 0.586 | |
| SEBackbone=Qwen2.5-7B-Instruct2026.01 | 0.544 | |
| LN-PEBackbone=Qwen2.5-7B-Instruct2026.01 | 0.541 | |
| SEBackbone=Qwen2.5-14B-Instruct2026.01 | 0.532 | |
| SEBackbone=Qwen2.5-32B-Instruct2026.01 | 0.528 | |
| LN-PEBackbone=Qwen2.5-14B-Instruct2026.01 | 0.526 | |
| LN-PEBackbone=Qwen2.5-32B-Instruct2026.01 | 0.524 |