Error Detection on TruthfulQA (full)
0.599AUROCB1 mean entropy
Evaluation Results
| Method | Links | |||
|---|---|---|---|---|
| B1 mean entropydescription=Baseline first-beam entropy2026.03 | 0.599 | 0.559 | — | |
| SelfCheck-EmbSample size (k)=52026.03 | 0.588 | 0.547 | 6 | |
| SE-JaccardSample size (N)=102026.03 | 0.548 | 0.51 | 11 | |
| SE-EmbeddingSample size (N)=102026.03 | 0.542 | 0.513 | 11 |