Reproduce “Learning Generative Dynamics for 3D Molecule Generation via Sequential Neuro-Symbolic Constraints” on OpenResearch
Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.
Universal app for Apple silicon and Intel. Requires macOS 11 or later.
About this paper
Zhuoran Li, Xu Sun, Chang Wen Chen, Wanyu Lin
SensDiff improves 3D molecule generation by training a denoiser with differentiable chemistry-rule penalties during refinement. It introduces rules in sequence, from preventing overlaps to guiding bond lengths and angles, then checking valence. On unconditional GEOM-Drugs generation, it reports 87.9% atom stability versus SLDM’s 84.8%; this benchmark result does not guarantee outputs are chemically valid or useful drug candidates.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.