List your company

Hallucination Detection on PopQA (AUPRC)

67.08AUPRCLyapunov Probes
7.823.1938.5853.97Mar 6, 2026

Evaluation results

MethodLinks
Lyapunov ProbesModel=Llama-3-8B2026.0367.08
ProbeModel=Qwen-3-4B2026.0364.41
Lyapunov ProbesModel=Qwen-3-4B2026.0364.02
Lyapunov ProbesModel=Llama-2-7B2026.0363.37
ProbeModel=Llama-2-7B2026.0361.87
Lyapunov ProbesModel=Falcon-7B2026.0361.23
ProbeModel=Llama-3-8B2026.0360.77
ProbeModel=Falcon-7B2026.0360.48
Seq. Prob.Model=Llama-3-8B2026.0327.02
VerbalizedModel=Llama-3-8B2026.0321.23
Seq. Prob.Model=Llama-2-7B2026.0320.22
VerbalizedModel=Llama-2-7B2026.0320.13
SurrogateModel=Llama-2-7B2026.0318.42
SurrogateModel=Llama-3-8B2026.0318.26
SurrogateModel=Falcon-7B2026.0316.53
Seq. Prob.Model=Falcon-7B2026.0315.13
Seq. Prob.Model=Qwen-3-4B2026.0313.79
VerbalizedModel=Falcon-7B2026.0313.76
VerbalizedModel=Qwen-3-4B2026.0312.04
SurrogateModel=Qwen-3-4B2026.0310.08