← back

Few-step Cofolding with All-Atom Flow Maps

Gianluca Scarpellini (et. al) · Paper · June 10, 2026

• Distilling protein cofolding and protein folding models seems like a super good idea. Maybe more the latter because cofolding models are kind of mid in comparison
• Nicholas Boffi’s semigroup prescription for learning flow maps is really good (https://arxiv.org/abs/2505.18825)
• I don’t understand the “Reparameterizing to Noise-Level Space” bit in the distillation process, I need to figure this out