Mathematical Reasoning on FormInv N=103 theorems v1
96.3AccuracyDeepSeek V3
Evaluation Results
| Method | Links | |||
|---|---|---|---|---|
| DeepSeek V3Coverage=≥98.6%2026.05 | 96.3 | 83.5 | 1.5 | |
| o1-miniCoverage=≥98.6%2026.05 | 95.2 | 79.6 | 0.4 | |
| Gemini 1.5F.Coverage=96.4%2026.05 | 94.2 | 82.5 | 2.5 | |
| Sonnet 3.5Coverage=≥98.6%2026.05 | 94 | 78.6 | 4.6 | |
| GPT-4oCoverage=≥98.6%2026.05 | 93.1 | 79.6 | 2.4 | |
| GPT-4o-miniCoverage=≥98.6%2026.05 | 92.6 | 73.8 | 4.2 | |
| DeepSeek R1Coverage=≥98.6%2026.05 | 91.9 | 75.7 | 2.3 | |
| Llama 70BCoverage=≥98.6%2026.05 | 89.8 | 67 | 1 | |
| Haiku 3.5Coverage=≥98.6%2026.05 | 85.7 | 54.4 | 4.4 |