General Reasoning on Reasoning GYM (Accuracy)
8.8Accuracy (Reasoning GYM)SAPO
Evaluation Results
| Method | Links | |
|---|---|---|
| SAPO2026.01 | 8.8 | |
| Pretrain + RL2026.01 | 8.6 | |
| GRPO2026.01 | 8.4 | |
| SpeedControl2026.01 | 7.4 | |
| CoScale-RL2026.01 | 6.8 | |
| COMPASS2026.01 | 6.8 | |
| Qwen2.5-0.5B Instruct2026.01 | 6 | |
| SFTed (w/o Data Scaling)Data Scaling=No2026.01 | 5.8 | |
| SFTed (w/ Data Scaling)Data Scaling=Yes2026.01 | 5.4 | |
| Scale-RL2026.01 | 5.4 |