Decode-gLM: A toolkit for interpreting, auditing, and steering genomic language models
Aaron Maiwald, Piotr Jedryszek, Florent Draye, Bernhard Schölkopf, Garrett M. Morris, Oliver M. Crook
This paper introduces Decode-gLM, a toolkit using sparse autoencoders to interpret, audit, and steer genomic language models, revealing biological features and identifying data contamination.
bioRxiv · Oct 31, 2025 · PreprintGenomics