Mathematical Reasoning on AMC 23 (Avg@16, #Token)
82.1Avg@16 ScoreDLER-R1-1.5B
Evaluation results
| Method | Links | ||
|---|---|---|---|
| DLER-R1-1.5BCategory=Length-oriented RL2026.03 | 82.1 | 2,419 | |
| GRPOCategory=Performance-oriented RL2026.03 | 81.9 | 9,917 | |
| GR³Category=Performance-oriented RL2026.03 | 81.6 | 4,153 | |
| Laser-DE-L4096-1.5BCategory=Length-oriented RL2026.03 | 73.4 | 3,110 | |
| DeepSeek-R1-Distill-1.5BCategory=Initial model2026.03 | 70.8 | 9,351 | |
| LCR1-1.5BCategory=Length-oriented RL2026.03 | 67.8 | 4,170 | |
| AdaptThink-1.5BCategory=Length-oriented RL2026.03 | 63.3 | 2,859 |