Relaxed Adsorption Energy Prediction on OC20-Dense (16k) (test)
0.382MAE (eV)QE-Catalytic-V2
Evaluation results
| Method | Links | ||||
|---|---|---|---|---|---|
| QE-Catalytic-V2Pretrain Data=OC20 (340k), Fine-tuning Data=OC20-Dense (16k), Inference Input=Text + 3D2026.05 | 0.382 | 0.847 | 46.4 | 19 | |
| QE-Catalytic-V2∆Pretrain Data=OC20 (340k), Fine-tuning Data=OC20-Dense (16k), Inference Input=3D structure (no textual description)2026.05 | 0.413 | 0.822 | 23.8 | 15.4 | |
| QE-CatalyticPretrain Data=OC20 (340k), Fine-tuning Data=OC20-Dense (16k)2026.05 | 0.427 | 0.818 | 21.2 | 14.9 | |
| QE-Catalytic-V2*Pretrain Data=OC20 (340k), Fine-tuning Data=OC20-Dense (16k), Inference Input=Text-only2026.05 | 0.464 | 0.802 | 14.3 | 12.6 | |
| GAP-CatBERTaPretrain Data=OC20 (340k), Fine-tuning Data=OC20-Dense (16k)2026.05 | 0.502 | 0.764 | 7.38 | 7.3 | |
| CatBERTaPretrain Data=None, Fine-tuning Data=OC20-Dense (16k)2026.05 | 0.542 | 0.712 | — | — |