Mathematical Reasoning on AIME 2024 (Accuracy, Length, Reduction)
57.5AccuracyOPSDC
Evaluation Results
| Method | Links | |||
|---|---|---|---|---|
| OPSDCBase Model=Qwen3-14B, Token Budget=8,192 tokens2026.03 | 57.5 | 4,599 | 37.4 | |
| OPSDCBase Model=Qwen3-8B, Token Budget=8,192 tokens2026.03 | 54.6 | 5,114 | 32.7 | |
| Concise promptBase Model=Qwen3-14B, Token Budget=8,192 tokens2026.03 | 45.4 | 6,483 | 11.7 | |
| Concise promptBase Model=Qwen3-8B, Token Budget=8,192 tokens2026.03 | 37.9 | 6,870 | 9.6 | |
| Base ModelBase Model=Qwen3-14B, Token Budget=8,192 tokens2026.03 | 27.1 | 7,345 | — | |
| Base ModelBase Model=Qwen3-8B, Token Budget=8,192 tokens2026.03 | 25 | 7,597 | — |