Output Prediction on CruxEval
94.13Pass@1DeepSeek-R1
Evaluation Results
| Method | Links | |||
|---|---|---|---|---|
| DeepSeek-R1Composition Level=L12026.04 | 94.13 | — | — | |
| QwQ-32BComposition Level=L12026.04 | 82.63 | — | — | |
| DeepSeek-R1Composition Level=L32026.04 | 79.25 | — | — | |
| DeepSeek-R1Composition Level=L22026.04 | 79.13 | — | — | |
| QwQ-32BComposition Level=L22026.04 | 74.5 | — | — | |
| QwQ-32BComposition Level=L32026.04 | 72.38 | — | — | |
| 1.8B Transformer PretrainTokens=50B, Decoding strategy=greedy decoding2026.03 | — | 34.4 | 44.9 | |
| 1.8B Transformer PretrainTokens=150 B, Decoding strategy=greedy decoding2026.03 | — | 48 | 57.7 | |
| 32B CWM FinetuneTokens=50B, Decoding strategy=greedy decoding2026.03 | — | 77.9 | 83.2 |