1 Million Chemically Recyclable Polymers Dataset Released

Researchers have published a dataset of 1 million synthetically realizable ring-opening polymerization (ROP) polymer structures, designed for chemical recyclability and critical applications. The dataset was generated using Virtual Forward Synthesis (VFS) and polymer expert language models, including polyBART and POLYT5. VFS applies known reactions to existing monomers, while polyBART explores its learned latent space and POLYT5 uses sequence-to-sequence generation. The structures were filtered through stringent chemical heuristics and automated validation pipelines to ensure novelty, validity, and data quality. The dataset is available on arXiv with the identifier 2608.00149.