Mathematical Reasoning on MATH Levels 3, 4, 5 (test)
90Overall AccuracyCPO
Evaluation results
| Method | Links | |||||
|---|---|---|---|---|---|---|
| CPOType=Ours2026.02 | 90 | 96 | 92 | 82 | — | |
| APEType=Automated Prompting2026.02 | 89.33 | 94 | 92 | 82 | — | |
| PromptBreederType=Automated Prompting2026.02 | 88.67 | 92 | 94 | 80 | — | |
| HumanType=Human Prompts2026.02 | 88.33 | 95 | 91 | 79 | — | |
| OPROType=Automated Prompting2026.02 | 88.33 | 94 | 92 | 79 | — | |
| PromptAgentType=Automated Prompting2026.02 | 88.33 | 95 | 93 | 77 | — | |
| CoT (1-shot)Type=Conventional Prompting2026.02 | 86.33 | 93 | 92 | 74 | — | |
| ReflexionType=Conventional Prompting2026.02 | 86.33 | 90 | 92 | 77 | — | |
| TextGradType=Automated Prompting2026.02 | 86 | 92 | 87 | 79 | — | |
| DSPyType=Automated Prompting2026.02 | 71.67 | 78 | 75 | 62 | — | |
| CyberCorrectBackbone=GPT-42026.05 | 59.6 | — | — | — | 7.3 | |
| CyberCorrectBackbone=Claude-3.52026.05 | 57.8 | — | — | — | 8.1 | |
| CoVeBackbone=GPT-42026.05 | 56.4 | — | — | — | 11.8 | |
| Self-RefineBackbone=GPT-42026.05 | 55.2 | — | — | — | 15.7 | |
| No-CorrectionBackbone=GPT-42026.05 | 51.8 | — | — | — | — |