Mathematical Reasoning on AIME 2024 (Pass@1, Pass@K)
12Pass@1GRPO + HORA
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| GRPO + HORAScale=7B2026.05 | 12 | 73.3 | |
| GRPOScale=7B2026.05 | 11.7 | 70 | |
| GRPOScale=3B2026.05 | 6.2 | 56.7 | |
| GRPO + HORAScale=3B2026.05 | 5.7 | 60 | |
| BaseScale=7B2026.05 | 4.8 | 70 | |
| GRPOScale=1.5B-Inst.2026.05 | 3.1 | 53.3 | |
| GRPO + HORAScale=1.5B-Inst.2026.05 | 3 | 63.3 | |
| BaseScale=1.5B-Inst.2026.05 | 2.2 | 66.7 | |
| BaseScale=3B2026.05 | 1.2 | 53.3 |