Model Ranking Prediction on Helpsteer 30B+ Models Holdout (test)
76.5Pairwise Accuracy (RM1)BENCHALIGN
Evaluation Results
| Method | Links | ||||
|---|---|---|---|---|---|
| BENCHALIGNHoldout=30B+ Models2026.02 | 76.5 | 63.7 | 0.71 | 0.387 | |
| TINYBENCHMARKSHoldout=30B+ Models2026.02 | 60.8 | 49.1 | 0.324 | -0.013 | |
| METABENCHHoldout=30B+ Models2026.02 | 60.6 | 49.7 | 0.328 | 0.007 | |
| RANDOMHoldout=30B+ Models2026.02 | 55.1 | 45.2 | 0.108 | -0.148 |