Input Prediction on CruxEval
66.5Pass@1 (inv_step_call)32B CWM Finetune
Evaluation Results
| Method | Links | |
|---|---|---|
| 32B CWM FinetuneTokens=50B, Decoding strategy=greedy decoding2026.03 | 66.5 | |
| 1.8B Transformer PretrainTokens=150 B, Decoding strategy=greedy decoding2026.03 | 53.6 | |
| 1.8B Transformer PretrainTokens=50B, Decoding strategy=greedy decoding2026.03 | 40.7 |