Uncertainty Calibration on SciBench
78.7AUROCPE
Evaluation Results
| Method | Links | |
|---|---|---|
| PEBackbone=Qwen2.5-7B-Instruct2026.01 | 78.7 | |
| ET-PEBackbone=Qwen2.5-7B-Instruct2026.01 | 78.5 | |
| p(True)Backbone=Qwen2.5-7B-Instruct2026.01 | 77.1 | |
| LSBackbone=Qwen2.5-32B-Instruct2026.01 | 76.8 | |
| ET-PEBackbone=Qwen2.5-32B-Instruct2026.01 | 76.7 | |
| LSBackbone=Qwen2.5-7B-Instruct2026.01 | 76.5 | |
| PEBackbone=Qwen2.5-32B-Instruct2026.01 | 76.5 | |
| LSBackbone=Qwen2.5-14B-Instruct2026.01 | 75.3 | |
| ET-PEBackbone=Qwen2.5-14B-Instruct2026.01 | 75.1 | |
| PEBackbone=Qwen2.5-14B-Instruct2026.01 | 73.7 | |
| p(True)Backbone=Qwen2.5-32B-Instruct2026.01 | 70.8 | |
| SEBackbone=Qwen2.5-14B-Instruct2026.01 | 63 | |
| SEBackbone=Qwen2.5-32B-Instruct2026.01 | 63 | |
| LN-PEBackbone=Qwen2.5-14B-Instruct2026.01 | 62.2 | |
| p(True)Backbone=Qwen2.5-14B-Instruct2026.01 | 62 | |
| LN-PEBackbone=Qwen2.5-7B-Instruct2026.01 | 58.9 | |
| SEBackbone=Qwen2.5-7B-Instruct2026.01 | 57.7 | |
| LN-PEBackbone=Qwen2.5-32B-Instruct2026.01 | 55.9 |