Mathematical Reasoning on GSM8K (Fine-tuning Accuracy n=1k, 5k, 7k)
76.4Accuracy (n=1k) on GSM8KBuffer-and-Reinforce
Evaluation Results
| Method | Links | ||||
|---|---|---|---|---|---|
| Buffer-and-Reinforce2026.05 | 76.4 | 76.6 | 77.1 | 76.7 | |
| SFT2026.05 | 71.3 | 71.6 | 71 | 71.3 |
| Method | Links | ||||
|---|---|---|---|---|---|
| Buffer-and-Reinforce2026.05 | 76.4 | 76.6 | 77.1 | 76.7 | |
| SFT2026.05 | 71.3 | 71.6 | 71 | 71.3 |