Scientific Design on SciDesignBench de novo domains v1
48PK/PD ScoreGemini 3.1 Pro
Evaluation Results
| Method | Links | ||||||||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Gemini 3.1 ProK=3, temperature=0.72026.03 | 48 | 31.7 | 1.7 | 13.3 | 23.3 | 70 | 15 | 18.3 | 8.3 | 31.7 | 23.3 | 0 | 28.3 | 24 | |
| Sonnet 4.5K=3, temperature=0.72026.03 | 38 | 53.3 | 0 | 20 | 3.3 | 50 | 48.3 | 33.3 | 1.7 | 30 | 30 | 45 | 26.7 | 29 | |
| GPT-5.2K=3, temperature=0.72026.03 | 38 | — | 18.3 | 18.3 | — | 55 | 33.3 | — | 0 | 30 | 13.3 | 31.7 | 18.3 | 25.6 | |
| Sonnet 4.6K=3, temperature=0.72026.03 | 29 | 46.7 | 1.7 | 16.7 | 20 | 63.3 | 26.7 | 28.3 | 0 | 30 | 0 | 28.3 | 33.3 | 22.9 | |
| Opus 4.6K=3, temperature=0.72026.03 | 23.3 | 56.7 | 3.3 | 13.3 | 5 | 53.3 | 33.3 | 28.3 | 3.3 | 33.3 | 25 | 30 | 20 | 23.8 | |
| Gemini 2.0 FlashK=3, temperature=0.72026.03 | 15 | 10 | 0 | 20 | 0 | 33.3 | 23.3 | 31.7 | 0 | 40 | 20 | 20 | 23.3 | 19.5 | |
| GPT-4oK=3, temperature=0.72026.03 | 0 | — | 20 | 18.3 | — | 16.7 | 0 | — | 0 | 35 | 11.7 | 18.3 | 8.3 | 12.8 |