Math Reasoning on Olympiad Math
1.26AccuracyCoScale-RL
Evaluation Results
| Method | Links | |
|---|---|---|
| CoScale-RL2026.01 | 1.26 | |
| COMPASS2026.01 | 0.72 | |
| SFTed (w/o Data Scaling)Data Scaling=No2026.01 | 0.51 | |
| SAPO2026.01 | 0.48 | |
| GRPO2026.01 | 0.46 | |
| Pretrain + RL2026.01 | 0.45 | |
| SpeedControl2026.01 | 0.45 | |
| Qwen2.5-0.5B Instruct2026.01 | 0.29 | |
| Scale-RL2026.01 | 0.29 | |
| SFTed (w/ Data Scaling)Data Scaling=Yes2026.01 | 0.24 |