Atomic Design Transformer: Scaffold-Conditioned 3D Molecule Generation with xTB-Verified Reinforcement Learning

Atomic Design Transformer: Scaffold-Conditioned 3D Molecule Generation with xTB-Verified Reinforcement Learning at https://arxiv.org/abs/2607.15918 source: https://github.com/tkotani/ADT (release v2.1) Dataset to reproduce the data in the paper.v2.1: the RLVR generator is rlvr_E240direct.pt (it replaces paper_rlvl.ckpt, which remains available in v2.0); frame_caches.tar.gz is corrected (every scaffold frame lays down exactly the named ring); geom_drugs_smiles.smi, the GEOM-Drugs SMILES list used for novelty and for the GEOM reference values, is added (derived from GEOM, Axelrod and Gómez-Bombarelli, Sci. Data 9, 185 (2022), CC0 1.0).To check the paper's numbers, see README_reproduce_benzene.md in the GitHub repository.