Hallucination Detection on SQuAD
0.721AUCsLSC
Evaluation Results
| Method | Links | ||||
|---|---|---|---|---|---|
| LSCModel=Qwen-0.5B2026.01 | 0.721 | — | 0.742 | 0.411 | |
| EigenScoreModel=Qwen-0.5B2026.01 | 0.698 | — | 0.712 | 0.366 | |
| Lexical SimilarityModel=Qwen-0.5B2026.01 | 0.67 | — | 0.698 | 0.317 | |
| AGSERModel=Qwen-0.5B2026.01 | 0.631 | — | 0.647 | 0.22 | |
| LN-EntropyModel=Qwen-0.5B2026.01 | 0.589 | — | 0.609 | 0.155 | |
| PerplexityModel=Qwen-0.5B2026.01 | 0.538 | — | 0.569 | 0.151 | |
| EigenScoreModel=Qwen-32B2026.01 | 0.491 | — | 0.512 | 0.076 | |
| LSCModel=Qwen-32B2026.01 | 0.483 | — | 0.517 | 0.081 | |
| EnergyModel=Qwen-0.5B2026.01 | 0.478 | — | 0.493 | -0.028 | |
| Lexical SimilarityModel=Qwen-32B2026.01 | 0.471 | — | 0.504 | 0.047 | |
| LN-EntropyModel=Qwen-32B2026.01 | 0.448 | — | 0.497 | 0.022 | |
| AGSERModel=Qwen-32B2026.01 | 0.425 | — | 0.492 | 0.027 | |
| PerplexityModel=Qwen-32B2026.01 | 0.345 | — | 0.443 | -0.063 | |
| EnergyModel=Qwen-32B2026.01 | 0.27 | — | 0.342 | -0.17 | |
| ICR ProbeModel=Qwen2.5-14B2025.07 | — | 0.7632 | — | — | |
| ICR ProbeModel=Qwen2.5-3B2025.07 | — | 0.7784 | — | — | |
| SAPLMAModel=Qwen2.5-14B2025.07 | — | 0.6984 | — | — | |
| SAPLMAModel=Qwen2.5-3B2025.07 | — | 0.6915 | — | — | |
| SEPModel=Qwen2.5-14B2025.07 | — | 0.6653 | — | — | |
| SEPModel=Qwen2.5-3B2025.07 | — | 0.656 | — | — |