NeurIPS 2026 Paper π-Fold Uses Physics-Informed RL to Fix Atom-Level Errors in Folding Models
lmthang · x · 2026-11-22
A repost introducing the team's first NeurIPS 2026 paper. The authors note that folding models nail the overall fold but not the atoms: trained to match the data distribution, they get the right shape with wrong atoms — clashing atoms and side chains in rotamers that don't exist.
The paper introduces π-Fold, which applies physics-informed reinforcement learning to improve folding models by bringing physical constraints into the training objective. A blog link is included.
More from Research
- Claude's System Card Reveals Protein-Sequence Training, Sparking Debate on Bio Foundation Models — chaitjo · 2026-11-22
- Watch neural-network cyclists evolve drafting and lane-change tactics in your browser — AjdDavison · 2026-11-22
- Axiom offers to pay biotechs to run drug programs while they keep 100% of assets — cjmaddison · 2026-11-22
- Cycling Neuroevolution: watch neural-net riders evolve racing tactics in your browser — AjdDavison · 2026-11-22
- AlphaProof Nexus settles open quantum optics question with Lean-verified proof — dair_ai · 2026-11-22
- Lean Community Verifies 17M Lines of Code, Confirms Claude-Assisted Riemann Hypothesis Proof — Polymarket · 2026-11-22