Changelog¶

v2.0.0 - 03/16/2026¶

New Features:

Adaptive per-transition probability threshold tuning to select the most informative transition-state samples
Multi-CPU parallel analysis via configurable n_workers parameter
Crash recovery with per-point checkpoint files; interrupted runs resume automatically
Forward feature selection using thermodynamics-inspired interpretation free energy
Built-in visualization functions: plot_feature_importance, plot_importance_heatmap, plot_unfaithfulness_curve, plot_point_variability
transition_summary and dominant_feature result analysis helpers
Support for numeric, angular, and sin/cos feature types
Configurable prob_threshold_min floor for probability thresholds
keep_checkpoints parameter to control per-point result file retention

Improvements:

Rewritten README with quick-start example, parameter table, and citation block
Expanded documentation with detailed usage guide, model script format, and visualization reference
Added CITATION.cff for standardized citation metadata

Improvement: