List your company

Hallucination Detection on Legal

97AUROCF: Ans. Expect.
41.8856.1970.584.81May 16, 2026

Evaluation results

MethodLinks
F: Ans. Expect.Backbone=Llama-3.3-70B2026.0597
TXTEMB (ctrl)Backbone=Llama-3.3-70B2026.0589
D: Act. VarianceBackbone=Llama-3.3-70B2026.0564
INSIDEBackbone=Llama-3.3-70B2026.0555
SelfCheckGPT-NLIBackbone=Llama-3.3-70B2026.0555
DRIFT-logpBackbone=Llama-3.3-70B2026.0554
PRISMBackbone=Llama-3.3-70B2026.0554
E: DRIFTBackbone=Llama-3.3-70B2026.0553
MINDBackbone=Llama-3.3-70B2026.0553
SEPsBackbone=Llama-3.3-70B2026.0553
DRIFT-concat (ablation of E)Backbone=Llama-3.3-70B2026.0552
Ensemble A+C+E+FBackbone=Llama-3.3-70B2026.0552
A: PerturbationBackbone=Llama-3.3-70B2026.0551
C: Sem. EntropyBackbone=Llama-3.3-70B2026.0550
IRIS⋆ (self-verify)Backbone=Llama-3.3-70B2026.0550
DoLaBackbone=Llama-3.3-70B2026.0550
Log-ProbBackbone=Llama-3.3-70B2026.0550
P(True)Backbone=Llama-3.3-70B2026.0550
HaMIBackbone=Llama-3.3-70B2026.0550
HalluShiftBackbone=Llama-3.3-70B2026.0549
ACT (detect)Backbone=Llama-3.3-70B2026.0549
HaloScopeBackbone=Llama-3.3-70B2026.0548
SAPLMABackbone=Llama-3.3-70B2026.0548
B: CAABackbone=Llama-3.3-70B2026.0544