Selective Generation on SamSum Out-of-domain
45.2PRR (ROUGE-L)Maximum Sequence Probability
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| Maximum Sequence ProbabilityModel=Llama 8b v3.1, Evaluation Setting=Out-of-domain, Training Protocol=Leave-one-out2025.02 | 45.2 | 21.8 | |
| HUQ-SATRMDModel=Llama 8b v3.1, Evaluation Setting=Out-of-domain, Training Protocol=Leave-one-out2025.02 | 41.3 | 22.9 | |
| Semantic EntropyModel=Llama 8b v3.1, Evaluation Setting=Out-of-domain, Training Protocol=Leave-one-out2025.02 | 31.1 | 20.6 | |
| SARModel=Llama 8b v3.1, Evaluation Setting=Out-of-domain, Training Protocol=Leave-one-out2025.02 | 30 | 21.7 | |
| DegMat NLI Score Entail.Model=Llama 8b v3.1, Evaluation Setting=Out-of-domain, Training Protocol=Leave-one-out2025.02 | 25 | 22.6 | |
| SAPLMAModel=Llama 8b v3.1, Evaluation Setting=Out-of-domain, Training Protocol=Leave-one-out2025.02 | 24.1 | 7.5 | |
| SATRMD+MSPModel=Llama 8b v3.1, Evaluation Setting=Out-of-domain, Training Protocol=Leave-one-out2025.02 | 20.7 | 31.4 | |
| FactoscopeModel=Llama 8b v3.1, Evaluation Setting=Out-of-domain, Training Protocol=Leave-one-out2025.02 | 6.1 | 3.7 |