Uncertainty Estimation on AmbigQA
78.5AUROCKernel Lang. Ent.
Evaluation Results
| Method | Links | |
|---|---|---|
| Kernel Lang. Ent.Backbone=Mistral-7B2026.04 | 78.5 | |
| TotalBackbone=Mistral-7B2026.04 | 76.8 | |
| SelfCheckGPTBackbone=Mistral-7B2026.04 | 73.3 | |
| AleatoricBackbone=Mistral-7B2026.04 | 71.6 | |
| Closeness CentralityBackbone=Mistral-7B2026.04 | 68.3 | |
| SemanticEntropyBackbone=Mistral-7B2026.04 | 67.8 | |
| SC + VCBackbone=Mistral-7B2026.04 | 67.1 | |
| PerplexityBackbone=Mistral-7B2026.04 | 66.1 | |
| SC Based VCBackbone=Mistral-7B2026.04 | 65.8 | |
| Max Token Prob.Backbone=Mistral-7B2026.04 | 65.8 | |
| Mean Token EntropyBackbone=Mistral-7B2026.04 | 65.4 | |
| Token EntropyBackbone=Mistral-7B2026.04 | 65.4 | |
| Max Sequence Prob.Backbone=Mistral-7B2026.04 | 65.1 | |
| SC ScoreBackbone=Mistral-7B2026.04 | 64.9 | |
| PTrueBackbone=Mistral-7B2026.04 | 60.4 | |
| Self CertaintyBackbone=Mistral-7B2026.04 | 56.9 |