Response-level Hallucination Detection on RAGTruth QA
91.89AUROCLettuceDetect-L
Evaluation Results
| Method | Links | |
|---|---|---|
| LettuceDetect-LDetector category=Black-box, Same-dataset evaluation=true2026.04 | 91.89 | |
| HDM-2-3BDetector category=Black-box, Same-dataset evaluation=true2026.04 | 87.95 | |
| LookbackLensDetector category=White-box, Proxy generator=Qwen3-4B-Instruct-2507, Same-dataset evaluation=true2026.04 | 87.54 | |
| RAGognizerDetector category=White-box, Proxy generator=Qwen3-4B-Instruct-25072026.04 | 87.43 | |
| HallucinationProbesDetector category=White-box, Probe=LoRA-KL probe on LLaMA-3.1-8B (lambda_KL = 0.05)2026.04 | 78.08 | |
| HHEM-2.1-OpenDetector category=Black-box2026.04 | 71.7 | |
| Semantic EntropyDetector category=White-box2026.04 | 67.98 | |
| MiniCheck-7BDetector category=Black-box2026.04 | 63.1 | |
| SelfCheckGPT (NLI)Detector category=Black-box2026.04 | 61.13 | |
| INSIDE (EigenScore)Detector category=White-box, Proxy generator=Qwen3-4B-Instruct-25072026.04 | 60.96 | |
| DeBERTa-v3 (Con.)Detector category=Black-box2026.04 | 56.69 | |
| HaloScopeDetector category=White-box, Proxy generator=Qwen3-4B-Instruct-2507, Same-dataset evaluation=true2026.04 | 53.54 | |
| DeBERTa-v3 (Ent.)Detector category=Black-box2026.04 | 53.25 |