Mathematical Reasoning on MATH (Avg@32 Accuracy)
87.71Avg@32 AccuracyRL
Evaluation results
| Method | Links | |
|---|---|---|
| RLBitwidth=BF162026.02 | 87.71 | |
| QuRL w/ UAQBitwidth=INT82026.02 | 87.2 | |
| QuRL w/o UAQBitwidth=INT82026.02 | 86.48 | |
| FlashRLBitwidth=INT82026.02 | 85.88 | |
| RLBitwidth=INT82026.02 | 84.9 | |
| BaseBitwidth=BF162026.02 | 82.9 |