Dataset for "Robust Steerability Classification via Key Feature Extraction"

This dataset accompanies the manuscript “Robust Steerability Classification via Key Feature Extraction” and contains the two-qubit quantum-state datasets used for training and evaluating machine-learning classifiers for quantum steerability. The collection covers projective-measurement settings (m=2,\ldots,8) and includes random two-qubit states, Werner states, T-diagonal states, strictly unsteerable random states, and all-versus-nothing (AVN) states. These datasets are used to evaluate both in-distribution classification performance and generalization across physically distinct families of quantum states. Each state is represented by the 15 independent real parameters of its density matrix together with a unified binary label, where (+1) denotes steerable and (-1) denotes either noncertified or unsteerable, depending on the dataset-specific certification criterion. The random-state datasets originate from Ren and Chen, Phys. Rev. A 100, 022314 (2019), while the remaining datasets are organized according to the constructions and steering criteria described in the accompanying manuscript. Detailed information on file structure, label definitions, state families, steering thresholds, and data provenance is provided in the included README.md. The dataset is released to support reproducibility of the numerical results reported in the manuscript and to provide benchmark data for future studies of machine-learning-based quantum-steering detection.