Math Reasoning on GSM8K (Coverage and Average Success Rate)
88CoverageTOFU
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| TOFUBackbone=QWEN-2.5-MATH-7B2026.04 | 88 | 46.8 | |
| CEBackbone=QWEN-2.5-MATH-7B2026.04 | 81.7 | 60.1 | |
| TOFUBackbone=QWEN-2.5-MATH-1.5B2026.04 | 80.6 | 51.8 | |
| CEBackbone=QWEN-2.5-MATH-1.5B2026.04 | 79.2 | 54.5 | |
| TOFUBackbone=DEEPSEEK-MATH-7B2026.04 | 79 | 47.5 | |
| GEMBackbone=QWEN-2.5-MATH-7B2026.04 | 76.2 | 61.5 | |
| CEBackbone=DEEPSEEK-MATH-7B2026.04 | 76 | 48 | |
| GEMBackbone=QWEN-2.5-MATH-1.5B2026.04 | 73.2 | 55.5 | |
| GEMBackbone=DEEPSEEK-MATH-7B2026.04 | 72.4 | 51.2 |