Math reasoning on MATH500 (Coverage and Success Rate)
86CoverageTOFU
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| TOFUBackbone=QWEN-2.5-MATH-7B2026.04 | 86 | 53.8 | |
| CEBackbone=QWEN-2.5-MATH-7B2026.04 | 84.8 | 54.2 | |
| GEMBackbone=QWEN-2.5-MATH-7B2026.04 | 83.2 | 59 | |
| TOFUBackbone=QWEN-2.5-MATH-1.5B2026.04 | 80.6 | 50.3 | |
| CEBackbone=QWEN-2.5-MATH-1.5B2026.04 | 78.4 | 53.9 | |
| GEMBackbone=QWEN-2.5-MATH-1.5B2026.04 | 78.4 | 53.4 | |
| TOFUBackbone=DEEPSEEK-MATH-7B2026.04 | 72.6 | 33 | |
| GEMBackbone=DEEPSEEK-MATH-7B2026.04 | 72.2 | 36.4 | |
| CEBackbone=DEEPSEEK-MATH-7B2026.04 | 71.4 | 32.7 |