Atomistic editing on ALM Bench (test)
61.3Efficacy (Ef)ALM Edit
Evaluation Results
| Method | Links | |||||||||
|---|---|---|---|---|---|---|---|---|---|---|
| ALM Edit2026.06 | 61.3 | 62.4 | 0.353 | 0.367 | 0.451 | 0.355 | 22.4 | 87.9 | 15.1 | |
| GPT-4othinking mode=enabled, prompting strategy=generate CIFs2026.06 | 50.5 | 46.9 | 0.024 | 0.127 | 0.081 | 0.018 | 4 | 0.7 | 0 | |
| GPT-4.1thinking mode=enabled, prompting strategy=generate CIFs2026.06 | 46.5 | 49.6 | 0.007 | 0.239 | 0.276 | 0.04 | 8.3 | 0.3 | 0 | |
| GPT-5.2thinking mode=enabled, prompting strategy=generate CIFs2026.06 | 43.7 | 41.4 | 0.058 | 0.244 | 0.006 | 0.032 | 11.8 | 0.2 | 0 |