Confidence Estimation on Internal Verification Suite GPT-4.1-mini (test)
0.912Q1 Attribution Present (AUROC)Logprob
Evaluation Results
| Method | Links | |||||||
|---|---|---|---|---|---|---|---|---|
| LogprobBase Model=GPT-4.1-mini2026.05 | 0.912 | 0.716 | 0.706 | 0.84 | 0.905 | 0.995 | 0.602 | |
| VerbalizedBase Model=GPT-4.1-mini2026.05 | 0.851 | 0.632 | — | 0.61 | — | — | 0.735 | |
| VERDI CVBase Model=GPT-4.1-mini, Evaluation Protocol=5-fold CV2026.05 | 0.806 | 0.825 | 0.915 | 0.877 | 0.905 | 1 | 0.459 |