Mathematical Reasoning on AIME 2024 (Pass@1 and Token Usage)
53.3Pass@1DyCon
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| DyConBackbone=R1-Llama-8B2026.06 | 53.3 | 11,876 | |
| BaselineBackbone=R1-Llama-8B2026.06 | 40 | 14,504 | |
| NoThinkingBackbone=R1-Llama-8B2026.06 | 36.7 | 9,447 |
| Method | Links | ||
|---|---|---|---|
| DyConBackbone=R1-Llama-8B2026.06 | 53.3 | 11,876 | |
| BaselineBackbone=R1-Llama-8B2026.06 | 40 | 14,504 | |
| NoThinkingBackbone=R1-Llama-8B2026.06 | 36.7 | 9,447 |