Home▸AI/ML Drug Discovery Infrastructure▸3D REINVENT-Style Augmented-Likelihood Policy for Atom Chains: Prior, Target Hetero-Atom Fraction and the Gibbs Bound

🧪 3D REINVENT-Style Augmented-Likelihood Policy for Atom Chains: Prior, Target Hetero-Atom Fraction and the Gibbs Bound

A Blender-built lab bench where a REINVENT-style generator learns to write 6-atom chains: marble towers show the share of each atom at each position against the frozen prior, and a rail of ball-and-bond chains shows the latest batch, with exact enumeration of all 15625 chains and the Gibbs variational bound.

AI/ML Drug Discovery Infrastructure3DAdvanced60 FPS📱 Mobile-adapted🟠 Blender⇄ 2D version
3D REINVENT-Style Augmented-Likelihood Policy for Atom Chains: Prior, Target Hetero-Atom Fraction and the Gibbs Bound ↗ Open standalone
⚙ Under the hood

A Blender-built lab bench where a REINVENT-style generator learns to write 6-atom chains: marble towers show the share of each atom at each position against the frozen prior, and a rail of ball-and-bond chains shows the latest batch, with exact enumeration of all 15625 chains and the Gibbs variational bound.

REINVENTPolicy GradientReinforcement LearningGenerative ChemistryAugmented LikelihoodKL DivergenceGibbs Variational PrincipleAI Drug Discovery Infrastructure

3D · Three.js / WebGL renderer · 60 FPS target · runs fully client-side, no install

What did you find?

Add reproduction steps (optional)