Model Ranking Prediction on Helpsteer 13B+ Models Holdout (test)
74.1Acc_pair (RM1 Helpful)BENCHALIGN
Evaluation Results
| Method | Links | ||||
|---|---|---|---|---|---|
| BENCHALIGNHoldout=13B+ Models2026.02 | 74.1 | 69.6 | 0.674 | 0.552 | |
| METABENCHHoldout=13B+ Models2026.02 | 58.4 | 53 | 0.243 | 0.089 | |
| TINYBENCHMARKSHoldout=13B+ Models2026.02 | 57 | 52 | 0.203 | 0.061 | |
| RANDOMHoldout=13B+ Models2026.02 | 55.8 | 53.7 | 0.125 | 0.083 |