Performance Prediction on HellaSwag (HS) 10k (test)
0.8MAEMetabench
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| MetabenchSelection=Best for val., Prediction=ability-IRT2025.10 | 0.8 | 0.974 | |
| DISCOSelection=High JSD, Prediction=Sig. + RF2025.10 | 0.86 | 0.972 | |
| DISCOSelection=High PDS, Prediction=Sig. + RF2025.10 | 1.01 | 0.984 | |
| tinyBenchmarksSelection=Anchor-corr, Prediction=gp-IRT2025.10 | 1.27 | 0.937 | |
| DISCOSelection=High PDS, Prediction=Sig. + kNN2025.10 | 1.32 | 0.956 | |
| Model signatureSelection=Random, Prediction=Sig. + RF2025.10 | 1.36 | 0.938 | |
| Model signatureSelection=Random, Prediction=Sig. + kNN2025.10 | 1.49 | 0.899 | |
| DISCOSelection=High JSD, Prediction=Sig. + kNN2025.10 | 1.5 | 0.944 | |
| tinyBenchmarksSelection=Random, Prediction=gp-IRT2025.10 | 1.96 | 0.819 | |
| tinyBenchmarksSelection=Anchor-IRT, Prediction=gp-IRT2025.10 | 2.19 | 0.83 | |
| BaselineSelection=Random, Prediction=Direct eval.2025.10 | 2.85 | 0.839 |