Hallucination Detection on FactualBio paragraph-length biographies
72.9AUROCHalluSE
Evaluation Results
| Method | Links | ||||
|---|---|---|---|---|---|
| HalluSEBackbone=Qwen3-32B, SHARS=applied: false2026.06 | 72.9 | 77.3 | 75.4 | 72.8 | |
| P(True)Backbone=Qwen3-32B2026.06 | 69.8 | 73.3 | 70 | 70 | |
| Naive SEBackbone=Qwen3-32B2026.06 | 66.2 | 73.1 | 70.5 | 70.5 | |
| Self-CheckBackbone=Qwen3-32B2026.06 | 57.6 | 69.3 | 73.5 | 73.5 |