Correctness Detection on Non-MAQA
77.53AUROCCredal
Evaluation results
| Method | Links | ||
|---|---|---|---|
| CredalBackbone=GPT-52026.03 | 77.53 | 2.67 | |
| ProbIntBackbone=GPT-52026.03 | 77.09 | 2.33 | |
| CoTBackbone=GPT-52026.03 | 76.95 | 5.33 | |
| Top-4Backbone=GPT-52026.03 | 76.68 | 3.33 | |
| VanillaBackbone=GPT-52026.03 | 75.77 | 4.17 | |
| ProbIntBackbone=gemini-2.5-pro2026.03 | 73.03 | 2.33 | |
| PosBackbone=GPT-52026.03 | 72.3 | 3.67 | |
| Top-4Backbone=gemini-2.5-pro2026.03 | 70.27 | 3.33 | |
| CredalBackbone=gemini-2.5-pro2026.03 | 69.78 | 2.67 | |
| VanillaBackbone=gemini-2.5-pro2026.03 | 68.89 | 4.17 | |
| PosBackbone=gemini-2.5-pro2026.03 | 68.06 | 3.67 | |
| MI-ClarificationsBackbone=gemini-2.5-pro2026.03 | 65.46 | 7.83 | |
| MI-ClarificationsBackbone=GPT-52026.03 | 65.28 | 7.83 | |
| Label prob.Backbone=gemini-2.5-pro2026.03 | 64.83 | 7.83 | |
| CoTBackbone=gemini-2.5-pro2026.03 | 64.31 | 5.33 | |
| Label prob.Backbone=GPT-52026.03 | 64.3 | 7.83 | |
| Is true prob.Backbone=gemini-2.5-pro2026.03 | 62.96 | 8.67 | |
| Is true prob.Backbone=GPT-52026.03 | 58.35 | 8.67 | |
| DiNCOBackbone=gemini-2.5-pro2026.03 | 57.49 | 9.17 | |
| DiNCOBackbone=GPT-52026.03 | 56.42 | 9.17 |