Physical Diagram Generation on PhyDrawGen Optics (Evaluation Set)
53.7Preference RatePhyDrawGen
Evaluation Results
| Method | Links | |
|---|---|---|
| PhyDrawGenEvaluation Protocol=Human-judge forced-choice comparison2026.05 | 53.7 | |
| GPT-5Evaluation Protocol=Human-judge forced-choice comparison2026.05 | 20.9 | |
| Gemini 3 ProEvaluation Protocol=Human-judge forced-choice comparison2026.05 | 17.8 | |
| Gemini 2.5 FlashEvaluation Protocol=Human-judge forced-choice comparison2026.05 | 7.6 |