Mathematical Reasoning on AIME 2025 (Accuracy and Token Costs)
40Sample CountSC
Evaluation Results
| Method | Links | ||||
|---|---|---|---|---|---|
| SCBackbone=Qwen2.5-3B2026.02 | 40 | — | 35.6 | 6.67 | |
| SCBackbone=Qwen2.5-7B2026.02 | 40 | — | 37.3 | 16.67 | |
| SCBackbone=Gemma3-4B2026.02 | 40 | — | 164.1 | 16.67 | |
| ESCBackbone=Qwen2.5-3B2026.02 | 39.83 | — | 37.8 | 6.67 | |
| DSCBackbone=Qwen2.5-3B2026.02 | 37.9 | 4.9 | 35.1 | 6.67 | |
| ESCBackbone=Qwen2.5-7B2026.02 | 37.64 | — | 37.4 | 16.67 | |
| ACBackbone=Qwen2.5-3B2026.02 | 36.73 | — | 42.4 | 6.67 | |
| DSCBackbone=Qwen2.5-7B2026.02 | 33.86 | 5.2 | 32.7 | 16.67 | |
| ESCBackbone=Gemma3-4B2026.02 | 32.72 | — | 136.3 | 16.67 | |
| ACBackbone=Qwen2.5-7B2026.02 | 32.58 | — | 39.3 | 16.67 | |
| ACTSCBackbone=Qwen2.5-3B2026.02 | 30.13 | — | 29.4 | 10 | |
| DSCBackbone=Gemma3-4B2026.02 | 28.96 | 17.7 | 119.8 | 16.67 | |
| ACBackbone=Gemma3-4B2026.02 | 27.12 | — | 119.1 | 16.67 | |
| ACTSCBackbone=Qwen2.5-7B2026.02 | 26.2 | — | 27.6 | 13.33 | |
| ACTSCBackbone=Gemma3-4B2026.02 | 17.9 | — | 74.5 | 16.67 |