The MSSPE-Dataset is a large-scale benchmark designed for the determination of stellar atmospheric parameters. It integrates high-resolution optical spectra from LAMOST, multi-band photometric images and magnitudes from SDSS and  WISE mission. 
The dataset comprises 240,436 samples, categorized into low-SNR (5≤SNR≤50; 96,275 samples, msspe_01.zip) and high-SNR (SNR > 50; 144,161 samples, msspe_02.zip) subsets.

All data are partitioned into training, validation, and test sets at a 7:1:2 ratio. 
While LAMOST spectra are consistently provided for all sources, the dataset incorporates real-world observational incompleteness in its photometric modalities. This dataset provides a structured basis for evaluating the accuracy and robustness of stellar parameter estimation across different noise levels and data availability scenarios.

(Detailed information regarding the file structures and contents of the msspe_01 and msspe_02 directories can be found in the README file. )