Hallucination Detection on Legal
97AUROCF: Ans. Expect.
Evaluation Results
| Method | Links | |
|---|---|---|
| F: Ans. Expect.Backbone=Llama-3.3-70B2026.05 | 97 | |
| TXTEMB (ctrl)Backbone=Llama-3.3-70B2026.05 | 89 | |
| D: Act. VarianceBackbone=Llama-3.3-70B2026.05 | 64 | |
| INSIDEBackbone=Llama-3.3-70B2026.05 | 55 | |
| SelfCheckGPT-NLIBackbone=Llama-3.3-70B2026.05 | 55 | |
| DRIFT-logpBackbone=Llama-3.3-70B2026.05 | 54 | |
| PRISMBackbone=Llama-3.3-70B2026.05 | 54 | |
| E: DRIFTBackbone=Llama-3.3-70B2026.05 | 53 | |
| MINDBackbone=Llama-3.3-70B2026.05 | 53 | |
| SEPsBackbone=Llama-3.3-70B2026.05 | 53 | |
| DRIFT-concat (ablation of E)Backbone=Llama-3.3-70B2026.05 | 52 | |
| Ensemble A+C+E+FBackbone=Llama-3.3-70B2026.05 | 52 | |
| A: PerturbationBackbone=Llama-3.3-70B2026.05 | 51 | |
| C: Sem. EntropyBackbone=Llama-3.3-70B2026.05 | 50 | |
| IRIS⋆ (self-verify)Backbone=Llama-3.3-70B2026.05 | 50 | |
| DoLaBackbone=Llama-3.3-70B2026.05 | 50 | |
| Log-ProbBackbone=Llama-3.3-70B2026.05 | 50 | |
| P(True)Backbone=Llama-3.3-70B2026.05 | 50 | |
| HaMIBackbone=Llama-3.3-70B2026.05 | 50 | |
| HalluShiftBackbone=Llama-3.3-70B2026.05 | 49 | |
| ACT (detect)Backbone=Llama-3.3-70B2026.05 | 49 | |
| HaloScopeBackbone=Llama-3.3-70B2026.05 | 48 | |
| SAPLMABackbone=Llama-3.3-70B2026.05 | 48 | |
| B: CAABackbone=Llama-3.3-70B2026.05 | 44 |