Reasoning on GSM8K (Diversity and Pass@K Metrics)
0.26Distinct-1GRPO-likelihood
Evaluation Results
| Method | Links | ||||||
|---|---|---|---|---|---|---|---|
| GRPO-likelihoodobjective=likelihood2025.09 | 0.26 | 0.86 | 0.53 | 0.59 | 50.9 | 80.4 | |
| GRPO-entropyobjective=entropy2025.09 | 0.1 | 0.38 | 0.12 | 0.25 | 77 | 92.6 | |
| DQO2025.09 | 0.1 | 0.42 | 0.14 | 0.31 | 76.3 | 91.2 | |
| GRPO2025.09 | 0.09 | 0.32 | 0.09 | 0.21 | 76.8 | 87.9 |