Generative geometric reasoning on GenExam
52.3Math-Rel ScoreFaire
Evaluation results
| Method | Links | ||
|---|---|---|---|
| Faire2026.03 | 52.3 | 9.3 | |
| GPT-Image-12026.03 | 52 | 8 | |
| Gemini-2.5-Image2026.03 | 43.1 | 0.7 | |
| Imagen-4-Ultra2026.03 | 35.9 | 2.6 | |
| Qwen-Image2026.03 | 18.9 | 0 | |
| Janus-Pro2026.03 | 13.7 | 0 | |
| Emu32026.03 | 11.3 | 0 |