Mathematical Reasoning on AIME 2025 (Acc, Tpi, Tok, Avg. Acc)
71.25AccuracyQwen3-32B-SCALE
Evaluation Results
| Method | Links | ||||
|---|---|---|---|---|---|
| Qwen3-32B-SCALEBackbone=Qwen3-32B, Strategy=SCALE2025.11 | 71.25 | 3,709 | 26,643 | 84.2 | |
| QwQ-SCALEBackbone=QwQ, Strategy=SCALE2025.11 | 70.21 | 4,673 | 33,965 | 81.95 | |
| Qwen3-32B-InftyThinkBackbone=Qwen3-32B, Strategy=InftyThink2025.11 | 70 | 6,538 | 36,640 | 83.96 | |
| QwQ-InftyThinkBackbone=QwQ, Strategy=InftyThink2025.11 | 65.53 | 10,114 | 73,473 | 79.99 | |
| QwQ-CoTBackbone=QwQ, Strategy=Chain-of-Thought2025.11 | 63.33 | 11,087 | 11,087 | 78.4 | |
| QwQ-Majority-VotingBackbone=QwQ, Strategy=Majority-Voting2025.11 | 63.33 | — | 88,696 | 79.17 | |
| Distill-Llama-70B-SCALEBackbone=Distill-Llama-70B, Strategy=SCALE2025.11 | 62.08 | 4,913 | 35,517 | 76.74 | |
| Distill-Llama-70B-CoTBackbone=Distill-Llama-70B, Strategy=Chain-of-Thought2025.11 | 61.37 | 8,218 | 8,218 | 75.56 | |
| Distill-Qwen-32B-Majority-VotingBackbone=Distill-Qwen-32B, Strategy=Majority-Voting2025.11 | 60 | — | 69,595 | 78.33 | |
| Distill-Qwen-32B-SCALEBackbone=Distill-Qwen-32B, Strategy=SCALE2025.11 | 59.17 | 4,140 | 29,641 | 76.04 | |
| Distill-Qwen-32B-CoTBackbone=Distill-Qwen-32B, Strategy=Chain-of-Thought2025.11 | 58.75 | 8,699 | 8,699 | 75.42 | |
| Qwen3-32B-CoTBackbone=Qwen3-32B, Strategy=Chain-of-Thought2025.11 | 57.5 | 6,839 | 6,839 | 75.9 | |
| Distill-Llama-70B-Majority-VotingBackbone=Distill-Llama-70B, Strategy=Majority-Voting2025.11 | 56.67 | — | 63,830 | 76.11 | |
| Distill-Llama-70B-InftyThinkBackbone=Distill-Llama-70B, Strategy=InftyThink2025.11 | 54.58 | 5,362 | 38,763 | 72.64 | |
| Qwen3-32B-Majority-VotingBackbone=Qwen3-32B, Strategy=Majority-Voting2025.11 | 53.33 | — | 54,719 | 75.83 | |
| Distill-Qwen-32B-InftyThinkBackbone=Distill-Qwen-32B, Strategy=InftyThink2025.11 | 48.75 | 4,915 | 35,187 | 69.69 |