Overall Performance on The benchmark held-out slice
85AccuracySolver-Grounded Synthetic CoT (Model)
Evaluation Results
| Method | Links | |
|---|---|---|
| Solver-Grounded Synthetic CoT (Model)adapter=best banked adapter2026.06 | 85 | |
| Base Model2026.06 | 46.6 |
| Method | Links | |
|---|---|---|
| Solver-Grounded Synthetic CoT (Model)adapter=best banked adapter2026.06 | 85 | |
| Base Model2026.06 | 46.6 |