Mathematical Reasoning on Math Benchmarks Overall (test)
87Pass@1SimPOShortest
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| SimPOShortestBackbone=Qwen3-8B2026.01 | 87 | 80.1 | |
| RELIEFBackbone=Qwen3-8B2026.01 | 86.1 | 71.3 | |
| BeliefPromptBackbone=Qwen3-8B2026.01 | 85.3 | 89 | |
| SFTbeliefBackbone=Qwen3-8B2026.01 | 84.8 | 95.4 | |
| OriginalBackbone=Qwen3-8B2026.01 | 84.7 | 100 | |
| SFTShortestBackbone=Qwen3-8B2026.01 | 83.9 | 98.8 | |
| RELIEFBackbone=DeepSeek-R1-Distill-Qwen-7B2026.01 | 75.3 | 89.7 | |
| SimPOShortestBackbone=DeepSeek-R1-Distill-Qwen-7B2026.01 | 74.7 | 91.5 | |
| OriginalBackbone=DeepSeek-R1-Distill-Qwen-7B2026.01 | 74 | 100 | |
| BeliefPromptBackbone=DeepSeek-R1-Distill-Qwen-7B2026.01 | 73.9 | 90.6 | |
| SFTShortestBackbone=DeepSeek-R1-Distill-Qwen-7B2026.01 | 71.7 | 85.1 | |
| SFTbeliefBackbone=DeepSeek-R1-Distill-Qwen-7B2026.01 | 71.1 | 90.5 |