Mathematical Reasoning on AIME 2025 (Acc %, Tokens)
86.67AccuracyRAG
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| RAGBackbone=Qwen3-14B2025.09 | 86.67 | 9,910.86 | |
| CoT-SCBackbone=Qwen3-4B2025.09 | 80 | 42,002.8 | |
| CoT-SCBackbone=Qwen3-8B2025.09 | 80 | 43,571.8 | |
| RoTBackbone=Qwen3-14B2025.09 | 80 | 10,187 | |
| CoTBackbone=Qwen3-14B2025.09 | 73.33 | 9,451.27 | |
| RoT + TIBackbone=Qwen3-4B2025.09 | 66.67 | 9,421 | |
| RAGBackbone=Qwen3-8B2025.09 | 66.67 | 11,777.46 | |
| CoT-SCBackbone=Qwen3-14B2025.09 | 66.67 | 39,272.87 | |
| RoT + TIBackbone=Qwen3-14B2025.09 | 66.67 | 9,767.33 | |
| RAGBackbone=Qwen3-4B2025.09 | 60 | 11,670.26 | |
| RoTBackbone=Qwen3-4B2025.09 | 60 | 11,634.6 | |
| CoTBackbone=Qwen3-8B2025.09 | 60 | 12,770.6 | |
| RoTBackbone=Qwen3-8B2025.09 | 60 | 12,831.67 | |
| CoTBackbone=Qwen3-4B2025.09 | 53.33 | 11,488.33 | |
| RoT + TIBackbone=Qwen3-8B2025.09 | 53.33 | 11,279.07 | |
| BoTBackbone=Qwen3-8B2025.09 | 33.33 | 59,970.86 | |
| CoTBackbone=Qwen3-1.7B2025.09 | 26.67 | 13,111.87 | |
| RAGBackbone=Qwen3-1.7B2025.09 | 26.67 | 13,280.8 | |
| BoTBackbone=Qwen3-4B2025.09 | 26.67 | 50,194.53 | |
| CoT-SCBackbone=Qwen3-1.7B2025.09 | 20 | 47,804.8 | |
| RoTBackbone=Qwen3-1.7B2025.09 | 20 | 12,791.73 | |
| RoT + TIBackbone=Qwen3-1.7B2025.09 | 20 | 8,568.47 | |
| RAGBackbone=Qwen3-0.6B2025.09 | 13.33 | 10,523.86 | |
| CoT-SCBackbone=Qwen3-0.6B2025.09 | 13.33 | 44,520.4 | |
| BoTBackbone=Qwen3-1.7B2025.09 | 13.33 | 33,657.47 | |
| CoTBackbone=Qwen3-0.6B2025.09 | 6.67 | 10,004.93 | |
| BoTBackbone=Qwen3-0.6B2025.09 | 6.67 | 38,792.17 | |
| RoTBackbone=Qwen3-0.6B2025.09 | 6.67 | 11,994 | |
| RoT + TIBackbone=Qwen3-0.6B2025.09 | 6.67 | 5,277.87 |