Mathematical Reasoning on AIME 25 (Acc, Rounds, #Samples)
57.5AccuracySC@32
Evaluation results
| Method | Links | |||
|---|---|---|---|---|
| SC@32Model=Qwen3-4B-Instruct-Thinking2026.06 | 57.5 | 1 | 32 | |
| ASCModel=Qwen3-4B-Instruct-Thinking2026.06 | 57.5 | 15 | 15 | |
| RL-guidedModel=Qwen3-4B-Instruct-Thinking2026.06 | 57.1 | 2.8 | 11 | |
| ESCModel=Qwen3-4B-Instruct-Thinking2026.06 | 57 | 4.2 | 20.8 | |
| RL-guidedModel=Qwen3-1.7B-Thinking2026.06 | 44.6 | 3.5 | 14 | |
| ASCModel=Qwen3-1.7B-Thinking2026.06 | 44.2 | 18.7 | 18.7 | |
| SC@32Model=Qwen3-1.7B-Thinking2026.06 | 44.1 | 1 | 32 | |
| ESCModel=Qwen3-1.7B-Thinking2026.06 | 44.1 | 4.6 | 22.8 | |
| SC@32Model=GPT-4.1-nano2026.06 | 33.5 | 1 | 32 | |
| ASCModel=GPT-4.1-nano2026.06 | 33.5 | 20.8 | 20.8 | |
| RL-guidedModel=GPT-4.1-nano2026.06 | 33.4 | 6.9 | 17.3 | |
| ESCModel=GPT-4.1-nano2026.06 | 33.2 | 5.2 | 26 | |
| SC@32Model=Qwen3-0.6B-Thinking2026.06 | 30.1 | 1 | 32 | |
| ASCModel=Qwen3-0.6B-Thinking2026.06 | 30.1 | 24.2 | 24.2 | |
| RL-guidedModel=Qwen3-0.6B-Thinking2026.06 | 30 | 5 | 19.9 | |
| ESCModel=Qwen3-0.6B-Thinking2026.06 | 29.8 | 5.8 | 28.9 |