Further exploration of the machine-learning-based nuclear mass table

The datasets used in this study include machine-learning–predicted nuclear binding energies, residual proton–neutron interactions (Vpn), and alpha-decay energies. The dataset for newly measured nuclear binding energies represents the differences between theoretical and experimental values, corresponding to the data shown in Fig. 1. The Vpn dataset contains the residual proton–neutron interaction results based on various nuclear mass models and their machine-learning–optimized counterparts, corresponding to Figs. 2 and 3. The alpha-decay energy dataset provides the predicted α-decay energies for nuclei with proton numbers 119 and 120 under multiple theoretical models, as shown in Fig. 5. Both the Vpn and alpha-decay energy datasets were derived using the formulas presented in this study.In addition, the dataset of alpha-decay energies for superheavy nuclei includes all other available data, divided into two ranges: Z = 89–118 and Z = 119–126. The Z = 89–118 subset contains the data shown in Fig. 4. All datasets are provided in .csv format and can be opened using Excel software.These data are of fundamental importance for advancing nuclear structure research. Precise measurements of binding energies and Vpn provide stringent constraints on nuclear mass models, contributing to improved nuclear structure models and a deeper understanding of nucleosynthesis processes in stars. Meanwhile, the newly measured alpha-decay energies play a key role in studying nuclear structure, the stability of superheavy nuclei, and the development of nuclear models.