Uncertainty Estimation on QASPER
0.722AUROCTotal
Evaluation Results
| Method | Links | |
|---|---|---|
| TotalBackbone=Mistral-7B2026.04 | 0.722 | |
| AleatoricBackbone=Mistral-7B2026.04 | 0.623 | |
| SelfCheckGPTBackbone=Mistral-7B2026.04 | 0.619 | |
| SC + VCBackbone=Mistral-7B2026.04 | 0.581 | |
| Closeness CentralityBackbone=Mistral-7B2026.04 | 0.579 | |
| Kernel Lang. Ent.Backbone=Mistral-7B2026.04 | 0.573 | |
| SC Based VCBackbone=Mistral-7B2026.04 | 0.568 | |
| SC ScoreBackbone=Mistral-7B2026.04 | 0.567 | |
| SemanticEntropyBackbone=Mistral-7B2026.04 | 0.514 | |
| Self CertaintyBackbone=Mistral-7B2026.04 | 0.484 | |
| Mean Token EntropyBackbone=Mistral-7B2026.04 | 0.481 | |
| Token EntropyBackbone=Mistral-7B2026.04 | 0.481 | |
| Max Sequence Prob.Backbone=Mistral-7B2026.04 | 0.48 | |
| Max Token Prob.Backbone=Mistral-7B2026.04 | 0.468 | |
| PerplexityBackbone=Mistral-7B2026.04 | 0.464 | |
| PTrueBackbone=Mistral-7B2026.04 | 0.423 |