Question Answering on HotpotQA (PR-AUC)
0.681PR-AUCCocoaPPL
Evaluation Results
| Method | Links | |
|---|---|---|
| CocoaPPLStrategy=beam-search, Backbone=Gemma 3 4B base2025.12 | 0.681 | |
| DissimilarityStrategy=multinomial-sampling, Backbone=Gemma 3 4B base2025.12 | 0.626 | |
| CocoaPPLStrategy=multinomial-sampling, Backbone=Gemma 3 4B base2025.12 | 0.625 | |
| DissimilarityStrategy=beam-search, Backbone=Gemma 3 4B base2025.12 | 0.611 | |
| CocoaMSPStrategy=beam-search, Backbone=Gemma 3 4B base2025.12 | 0.523 | |
| EccentricityStrategy=beam-search, Backbone=Gemma 3 4B base2025.12 | 0.514 | |
| EigVecDissimilarityStrategy=beam-search, Backbone=Gemma 3 4B base2025.12 | 0.487 | |
| EccentricityStrategy=multinomial-sampling, Backbone=Gemma 3 4B base2025.12 | 0.453 | |
| EigVecDissimilarityStrategy=multinomial-sampling, Backbone=Gemma 3 4B base2025.12 | 0.448 | |
| CocoaMSPStrategy=multinomial-sampling, Backbone=Gemma 3 4B base2025.12 | 0.43 | |
| Degree MatrixStrategy=multinomial-sampling, Backbone=Gemma 3 4B base2025.12 | 0.311 | |
| SARStrategy=multinomial-sampling, Backbone=Gemma 3 4B base2025.12 | 0.296 | |
| Semantic EntropyStrategy=beam-search, Backbone=Gemma 3 4B base2025.12 | 0.278 | |
| SARStrategy=beam-search, Backbone=Gemma 3 4B base2025.12 | 0.266 | |
| Degree MatrixStrategy=beam-search, Backbone=Gemma 3 4B base2025.12 | 0.244 | |
| Semantic EntropyStrategy=multinomial-sampling, Backbone=Gemma 3 4B base2025.12 | 0.14 |