DiSE: A multimodal model for automatic structure elucidation of organic compounds

This repository contains the pre-trained model checkpoints and datasets for DiSE (Diffusion-based Structure Elucidation), an end-to-end generative model designed for the automatic structure elucidation of organic compounds. For full methodological details and results, please refer to the preprint: arXiv:2510.26231 The source code will be released at: https://github.com/Anan-Wu-XMU/DiSE Files Included epoch=1173.ckpt, epoch=639.ckpt Pre-trained model weights (checkpoints) for the DiSE model (best model based on PCQM4MV2-NMR). QM9NMR_all_mol_label_v4.pickle A processed dataset containing molecular labels and associated NMR data derived from the QM9 dataset, used for training and evaluation.