Model Ranking Prediction on SuperGPQA
0.88Spearman Rank Correlation (rho)RankSVM (linear)
Evaluation Results
| Method | Links | |
|---|---|---|
| RankSVM (linear)Protocol=Leave-2-tasks-out cross-validated, Subsampling Seeds=202026.05 | 0.88 | |
| 3-sparse proxyProtocol=Leave-2-tasks-out cross-validated, Subsampling Seeds=202026.05 | 0.84 | |
| RankSVM (RBF)Protocol=Leave-2-tasks-out cross-validated, Subsampling Seeds=202026.05 | 0.84 | |
| Univariate proxyProtocol=Leave-2-tasks-out cross-validated, Subsampling Seeds=202026.05 | 0.64 | |
| rBridgeProtocol=Leave-2-tasks-out cross-validated2026.05 | 0.55 | |
| FineWeb CE lossProtocol=Leave-2-tasks-out cross-validated2026.05 | 0.54 |