Mathematical Reasoning on MATH 500 (Acc, Len, CR)
93.6AccuracyLCPO
Evaluation results
| Method | Links | |||
|---|---|---|---|---|
| LCPOBackbone=Qwen3-8B, Decoding Strategy=greedy decoding (pass@1)2026.04 | 93.6 | 4.8 | 27.8 | |
| ETRBackbone=Qwen3-8B, Decoding Strategy=greedy decoding (pass@1)2026.04 | 93.6 | 2.4 | 44.4 | |
| DEERBackbone=Qwen3-8B, Decoding Strategy=greedy decoding (pass@1)2026.04 | 93 | 3.1 | 57.4 | |
| PEARBackbone=Qwen3-8B, Decoding Strategy=greedy decoding (pass@1)2026.04 | 92.4 | 4.5 | 83.5 | |
| DEERBackbone=Qwen3-4B, Decoding Strategy=greedy decoding (pass@1)2026.04 | 92.2 | 3.7 | 74 | |
| O1-PrunerBackbone=Qwen3-8B, Decoding Strategy=greedy decoding (pass@1)2026.04 | 92.2 | 1.9 | 35.2 | |
| PEARBackbone=Qwen3-4B, Decoding Strategy=greedy decoding (pass@1)2026.04 | 91.8 | 3.8 | 36 | |
| ETRBackbone=Qwen3-4B, Decoding Strategy=greedy decoding (pass@1)2026.04 | 91.4 | 2.1 | 42 | |
| LCPOBackbone=Qwen3-4B, Decoding Strategy=greedy decoding (pass@1)2026.04 | 90.8 | 4.5 | 89.2 | |
| ETRBackbone=DeepSeek-R1-Distill-7B, Decoding Strategy=greedy decoding (pass@1)2026.04 | 90.6 | 1.5 | 35.7 | |
| OriginalBackbone=Qwen3-4B, Decoding Strategy=greedy decoding (pass@1)2026.04 | 90.6 | 5 | 100 | |
| O1-PrunerBackbone=Qwen3-4B, Decoding Strategy=greedy decoding (pass@1)2026.04 | 90.6 | 1.8 | 36 | |
| OriginalBackbone=Qwen3-8B, Decoding Strategy=greedy decoding (pass@1)2026.04 | 90.6 | 5.4 | 100 | |
| PEARBackbone=DeepSeek-R1-Distill-7B, Decoding Strategy=greedy decoding (pass@1)2026.04 | 90.2 | 2.5 | 59.9 | |
| O1-PrunerBackbone=DeepSeek-R1-Distill-7B, Decoding Strategy=greedy decoding (pass@1)2026.04 | 89 | 2.1 | 50 | |
| NoThinkBackbone=Qwen3-8B, Decoding Strategy=greedy decoding (pass@1)2026.04 | 86.4 | 1.5 | 27.8 | |
| DEERBackbone=DeepSeek-R1-Distill-7B, Decoding Strategy=greedy decoding (pass@1)2026.04 | 85.8 | 2.3 | 54.8 | |
| NoThinkBackbone=Qwen3-4B, Decoding Strategy=greedy decoding (pass@1)2026.04 | 85.6 | 1.7 | 34 | |
| LCPOBackbone=DeepSeek-R1-Distill-7B, Decoding Strategy=greedy decoding (pass@1)2026.04 | 85.4 | 2.2 | 52.6 | |
| OriginalBackbone=DeepSeek-R1-Distill-7B, Decoding Strategy=greedy decoding (pass@1)2026.04 | 85 | 4.2 | 100 | |
| NoThinkBackbone=DeepSeek-R1-Distill-7B, Decoding Strategy=greedy decoding (pass@1)2026.04 | 81.4 | 1.7 | 40.5 |