Competition-style Proof and Derivation on AIME 21-24
86.2AccuracyRaw CoT
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| Raw CoTLLM=DEEPSEEK-R1-0528-QWEN3-8B2026.06 | 86.2 | 8,190 | |
| CLSRLLM=DEEPSEEK-R1-0528-QWEN3-8B2026.06 | 82.5 | 2,979 | |
| SoTLLM=DEEPSEEK-R1-0528-QWEN3-8B2026.06 | 81.3 | 3,266 | |
| CCoTLLM=DEEPSEEK-R1-0528-QWEN3-8B2026.06 | 79.6 | 3,946 | |
| CoDLLM=DEEPSEEK-R1-0528-QWEN3-8B2026.06 | 77.4 | 3,545 | |
| Raw CoTLLM=QWEN3-32B2026.06 | 46.8 | 4,572 | |
| Raw CoTLLM=QWEN3-8B2026.06 | 46.1 | 4,234 | |
| CLSRLLM=QWEN3-32B2026.06 | 45.2 | 2,038 | |
| CLSRLLM=QWEN3-8B2026.06 | 43.7 | 2,361 | |
| SoTLLM=QWEN3-32B2026.06 | 38.5 | 2,453 | |
| SoTLLM=QWEN3-8B2026.06 | 38 | 2,637 | |
| CCoTLLM=QWEN3-32B2026.06 | 37.3 | 3,025 | |
| CCoTLLM=QWEN3-8B2026.06 | 36.7 | 2,805 | |
| CoDLLM=QWEN3-32B2026.06 | 32.5 | 2,912 | |
| CoDLLM=QWEN3-8B2026.06 | 30.2 | 2,610 | |
| CLSRLLM=LLaMA3-8B2026.06 | 6.2 | 1,035 | |
| Raw CoTLLM=LLaMA3-8B2026.06 | 6.1 | 1,705 | |
| CCoTLLM=LLaMA3-8B2026.06 | 5.8 | 1,205 | |
| SoTLLM=LLaMA3-8B2026.06 | 5.3 | 1,051 | |
| CoDLLM=LLaMA3-8B2026.06 | 4.9 | 1,041 |