Step-level Hallucination Detection on TruthfulQA
0.965AUROCStudent
Evaluation results
| Method | Links | |
|---|---|---|
| StudentDeployability=deployable2026.05 | 0.965 | |
| TeacherDeployability=non-deployable2026.05 | 0.96 | |
| Linear Probe2026.05 | 0.9 | |
| LLM-CheckMechanism=attention2026.05 | 0.698 | |
| TL-Perplexity2026.05 | 0.671 | |
| TL-Entropy2026.05 | 0.644 |