Mathematical Reasoning on AMC23 (Pass@1, Pass@K)
53.9Pass@1GRPO + HORA
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| GRPO + HORAScale=7B2026.05 | 53.9 | 97.5 | |
| GRPOScale=7B2026.05 | 49.8 | 100 | |
| GRPO + HORAScale=3B2026.05 | 36.3 | 97.5 | |
| GRPOScale=3B2026.05 | 35.9 | 97.5 | |
| GRPO + HORAScale=1.5B-Inst.2026.05 | 27.8 | 95 | |
| GRPOScale=1.5B-Inst.2026.05 | 27.4 | 87.5 | |
| BaseScale=7B2026.05 | 24.5 | 97.5 | |
| BaseScale=1.5B-Inst.2026.05 | 22 | 92.5 | |
| BaseScale=3B2026.05 | 11.1 | 100 |