Graph Neural Force Field Prediction on MD22 Ac-Ala3+AT-AT
1.245Force RMSEJoint FT
Evaluation results
| Method | Links | ||
|---|---|---|---|
| Joint FTStrategy=Joint Fine-Tuning, Precision=Full precision2026.06 | 1.245 | 0.0403 | |
| GFFMERGEMerging Strategy=GFFMERGE, Precision=Full precision2026.06 | 1.6522 | 0.0528 | |
| Joint FTBackbone=M3GNet2026.06 | 1.84 | 0.04 | |
| GFFMergeBackbone=M3GNet2026.06 | 2.32 | 0.06 | |
| Joint FT SBBackbone=M3GNet2026.06 | 2.85 | 0.07 | |
| Weight AveragingMerging Strategy=Weight Averaging, Precision=Full precision2026.06 | 2.8673 | 1.9555 | |
| Wt AvgBackbone=M3GNet2026.06 | 4.13 | 1.97 | |
| FisherBackbone=M3GNet2026.06 | 4.65 | 2.51 | |
| EMRBackbone=M3GNet2026.06 | 4.8 | 1.67 | |
| TIESBackbone=M3GNet2026.06 | 7.47 | 5.03 |