This release provides the first public version of our preprint Explaining What Matters: Faithfulness in Molecular Deep Learning
📄 Paper
We introduce a systematic, fidelity-based framework to check whether feature attribution methods truly reflect a model’s internal reasoning.