Sentence-Level Confidence Prediction on WikiBio
68.6AUROCStructural Signal
Evaluation results
| Method | Links | ||
|---|---|---|---|
| Structural Signal2026.02 | 68.6 | 45.4 | |
| Semantic-feature2026.02 | 68.3 | 43 | |
| KMeans-outlier2026.02 | 59.4 | 35.5 | |
| Local Variation2026.02 | 56.5 | 32 | |
| SelfCheckGPT2026.02 | 56.4 | 31.2 | |
| Structure-feature2026.02 | 54.3 | 35.4 | |
| Shape Coherence2026.02 | 54 | 29.5 | |
| Spectral Stability2026.02 | 50.4 | 30.9 | |
| Mean logp2026.02 | 48.5 | 29.5 | |
| Semantic entropyProtocol=NLI, single-pass2026.02 | 45.3 | 26.4 |