Hallucination Detection on CoQA (AUPRC)
89.01AUPRCLyapunov Probes
Evaluation Results
| Method | Links | |
|---|---|---|
| Lyapunov ProbesModel=Qwen-3-4B2026.03 | 89.01 | |
| ProbeModel=Qwen-3-4B2026.03 | 88.3 | |
| Lyapunov ProbesModel=Llama-3-8B2026.03 | 81.28 | |
| ProbeModel=Llama-3-8B2026.03 | 80.67 | |
| Lyapunov ProbesModel=Llama-2-7B2026.03 | 76.13 | |
| ProbeModel=Llama-2-7B2026.03 | 72.25 | |
| Lyapunov ProbesModel=Falcon-7B2026.03 | 66.03 | |
| ProbeModel=Falcon-7B2026.03 | 65.36 | |
| VerbalizedModel=Qwen-3-4B2026.03 | 62.56 | |
| Seq. Prob.Model=Qwen-3-4B2026.03 | 61.13 | |
| SurrogateModel=Qwen-3-4B2026.03 | 59.54 | |
| SurrogateModel=Llama-3-8B2026.03 | 53.5 | |
| SurrogateModel=Llama-2-7B2026.03 | 53.29 | |
| VerbalizedModel=Llama-3-8B2026.03 | 52.4 | |
| VerbalizedModel=Llama-2-7B2026.03 | 51.52 | |
| Seq. Prob.Model=Llama-3-8B2026.03 | 50.35 | |
| Seq. Prob.Model=Llama-2-7B2026.03 | 49.63 | |
| Seq. Prob.Model=Falcon-7B2026.03 | 37.13 | |
| VerbalizedModel=Falcon-7B2026.03 | 35.54 | |
| SurrogateModel=Falcon-7B2026.03 | 34.83 |