Mathematical Reasoning on AIME Short context 2022-2024
26.67Accuracy0.7 × RoPE theta × 16 + 0.3 × 1M-Model
Evaluation results
| Method | Links | ||
|---|---|---|---|
| 0.7 × RoPE theta × 16 + 0.3 × 1M-ModelOperation=0.7 × RoPE theta × 16 + 0.3 × 1M-Model, recipe=LCTMerge2025.05 | 26.67 | 20,265 | |
| RoPE theta × 16Operation=RoPE theta × 162025.05 | 25.78 | 15,036 | |
| RoPE theta × 1Operation=RoPE theta × 12025.05 | 16.22 | 3,463 |