Performance Prediction on Winogrande (WG) 1.3k (test)
1MAEDISCO
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| DISCOSelection=High PDS, Prediction=Sig. + RF2025.10 | 1 | 0.967 | |
| DISCOSelection=High JSD, Prediction=Sig. + RF2025.10 | 1.09 | 0.973 | |
| DISCOSelection=High PDS, Prediction=Sig. + kNN2025.10 | 1.19 | 0.951 | |
| MetabenchSelection=Best for val., Prediction=ability-IRT2025.10 | 1.23 | 0.947 | |
| DISCOSelection=High JSD, Prediction=Sig. + kNN2025.10 | 1.26 | 0.955 | |
| Model signatureSelection=Random, Prediction=Sig. + RF2025.10 | 1.29 | 0.926 | |
| Model signatureSelection=Random, Prediction=Sig. + kNN2025.10 | 1.58 | 0.92 | |
| tinyBenchmarksSelection=Random, Prediction=gp-IRT2025.10 | 1.64 | 0.928 | |
| tinyBenchmarksSelection=Anchor-corr, Prediction=gp-IRT2025.10 | 1.95 | 0.918 | |
| tinyBenchmarksSelection=Anchor-IRT, Prediction=gp-IRT2025.10 | 2.24 | 0.85 | |
| BaselineSelection=Random, Prediction=Direct eval.2025.10 | 3.6 | 0.827 |