Correctness Detection on Non-AmbigQA
79.32AUROCCredal
Evaluation results
| Method | Links | ||
|---|---|---|---|
| CredalBackbone=GPT-52026.03 | 79.32 | 2.67 | |
| VanillaBackbone=GPT-52026.03 | 77.56 | 4.17 | |
| CoTBackbone=GPT-52026.03 | 77.24 | 5.33 | |
| ProbIntBackbone=GPT-52026.03 | 77.13 | 2.33 | |
| Top-4Backbone=GPT-52026.03 | 76.66 | 3.33 | |
| PosBackbone=GPT-52026.03 | 74.26 | 3.67 | |
| ProbIntBackbone=gemini-2.5-pro2026.03 | 71.28 | 2.33 | |
| PosBackbone=gemini-2.5-pro2026.03 | 67.8 | 3.67 | |
| Top-4Backbone=gemini-2.5-pro2026.03 | 66.98 | 3.33 | |
| VanillaBackbone=gemini-2.5-pro2026.03 | 66.88 | 4.17 | |
| CoTBackbone=gemini-2.5-pro2026.03 | 60.94 | 5.33 | |
| CredalBackbone=gemini-2.5-pro2026.03 | 59.82 | 2.67 | |
| Is true prob.Backbone=gemini-2.5-pro2026.03 | 58.97 | 8.67 | |
| Label prob.Backbone=GPT-52026.03 | 58.76 | 7.83 | |
| MI-ClarificationsBackbone=GPT-52026.03 | 58.4 | 7.83 | |
| Is true prob.Backbone=GPT-52026.03 | 55.84 | 8.67 | |
| Label prob.Backbone=gemini-2.5-pro2026.03 | 55.82 | 7.83 | |
| MI-ClarificationsBackbone=gemini-2.5-pro2026.03 | 54.97 | 7.83 | |
| DiNCOBackbone=gemini-2.5-pro2026.03 | 52.75 | 9.17 | |
| DiNCOBackbone=GPT-52026.03 | 52.19 | 9.17 |