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.

JanusDNA, the first bidirectional DNA foundation model built upon a novel pretraining paradigm, integrating the optimization efficiency of autoregressive modeling with the bidirectional comprehension capability of masked modeling.

PatternLocal is a novel XAI method that refines local linearization approaches to reduce false-positive feature attributions in non-linear explanations.