Mathematical Problem Solving on AIME, AMC, MATH, GSM8K (test)
58.5AIME Accuracyp*
Evaluation Results
| Method | Links | |||||
|---|---|---|---|---|---|---|
| p*Setting=GRPO, Backbone=8B-Base, Evaluation Protocol=pass@1/mean@322026.04 | 58.5 | 74.7 | 86.7 | 95.2 | 78.8 | |
| p* smooth(alpha = 0.01)Setting=GRPO, Backbone=8B-Base, Evaluation Protocol=pass@1/mean@322026.04 | 58.5 | 74 | 86.7 | 94.9 | 78.5 | |
| No samplingSetting=GRPO, Backbone=8B-Base, Evaluation Protocol=pass@1/mean@322026.04 | 58.3 | 72.8 | 87 | 96 | 78.5 |