Replication Data for A simplified machine learning workflow for identifying potential singlet fission candidates: benzannulated biphenylenes as a case study

This dataset contains all the necessary information to reproduce the results presented in the manuscript "Streamlined Machine Learning Protocol for the Discovery of Singlet Fission Materials". It includes a multi-XYZ file with the optimized geometries of all the molecular structures studied, as well as accompanying .csv files that provide the corresponding SMILES strings. The target values were obtained through DFT and TD-DFT calculations performed with Gaussian 16, and molecular descriptors for all systems were generated using AQME. In addition to the computed properties, the dataset contains the predicted values produced by ROBERT, a tool for automating and documenting predictive models in computational chemistry, making it easier to bridge chemical research with modern machine learning techniques. To further support reproducibility and traceability, the dataset also include a PDF report generated by ROBERT, which document two successive rounds of model training and evaluation.