Hallucination Detection on CovidQA, PubmedQA, FinanceBench
82.3F1 ScoreHalluClean
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| HalluCleanBackbone=GPT-3.5-turbo2025.11 | 82.3 | 83.2 | |
| DeepSeek-R12025.11 | 64.2 | 72.5 | |
| DeepSeek-V32025.11 | 57.9 | 72.2 | |
| GPT-4o-mini2025.11 | 47.1 | 64 | |
| Llama-3-70B2025.11 | 10.5 | 52.2 | |
| GPT-3.5-turbo2025.11 | 9.4 | 52 |