MS student, Technische Universität Berlin
1 paper at NeurIPS 2025
We provide a generalisation of activation pattern-based approaches to correct the interpretations of additive models, particularly in the presence of suppressors and correlated noise.
