Correctness Prediction on GSM8K (test)
0.852Best-layer ROC-AUCTrajectory features
Evaluation Results
| Method | Links | |
|---|---|---|
| Trajectory features2026.04 | 0.852 | |
| LogitLensconfig=best config2026.04 | 0.765 | |
| Step-count only2026.04 | 0.649 |
Loading the SOTA2 catalog…
| Method | Links | |
|---|---|---|
| Trajectory features2026.04 | 0.852 | |
| LogitLensconfig=best config2026.04 | 0.765 | |
| Step-count only2026.04 | 0.649 |