RFD4's Latent Space Trained on Physics-Based Data, Merging Force Fields and Folding Models
The RFD4-Proteina team revealed that its latent space is physics-informed, trained on synthetic physics-based PDB data and OMol DFT molecular data with TMol (Rosetta force field) supervision, a merge of force fields and folding models.
2026-10-22 ~ 2026-10-22 · 2 related posts
- Force fields and folding models are merging: physics-informed latent spaces — TimothyDuignan · 2026-10-22
- RFD4's physics-informed latent space trained via force-field backprop through TMol — AllThingsApx · 2026-10-22