Correctness Detection on MMLU-Pro
87.98AUROCCredal
Evaluation results
| Method | Links | ||
|---|---|---|---|
| CredalBackbone=GPT-52026.03 | 87.98 | 2.67 | |
| PosBackbone=GPT-52026.03 | 87.38 | 3.67 | |
| Top-4Backbone=GPT-52026.03 | 87.25 | 3.33 | |
| ProbIntBackbone=GPT-52026.03 | 86.17 | 2.33 | |
| VanillaBackbone=GPT-52026.03 | 85.87 | 4.17 | |
| CoTBackbone=GPT-52026.03 | 85.48 | 5.33 | |
| PosBackbone=gemini-2.5-pro2026.03 | 78.63 | 3.67 | |
| ProbIntBackbone=gemini-2.5-pro2026.03 | 78.57 | 2.33 | |
| MI-ClarificationsBackbone=GPT-52026.03 | 77.9 | 7.83 | |
| Top-4Backbone=gemini-2.5-pro2026.03 | 76.52 | 3.33 | |
| Label prob.Backbone=GPT-52026.03 | 76.49 | 7.83 | |
| CredalBackbone=gemini-2.5-pro2026.03 | 74.66 | 2.67 | |
| VanillaBackbone=gemini-2.5-pro2026.03 | 73.26 | 4.17 | |
| DiNCOBackbone=gemini-2.5-pro2026.03 | 71.33 | 9.17 | |
| DiNCOBackbone=GPT-52026.03 | 70.87 | 9.17 | |
| Is true prob.Backbone=GPT-52026.03 | 68.5 | 8.67 | |
| CoTBackbone=gemini-2.5-pro2026.03 | 68.26 | 5.33 | |
| Is true prob.Backbone=gemini-2.5-pro2026.03 | 65.83 | 8.67 | |
| Label prob.Backbone=gemini-2.5-pro2026.03 | 63.77 | 7.83 | |
| MI-ClarificationsBackbone=gemini-2.5-pro2026.03 | 63.48 | 7.83 |