Do protein language models store different structural elements in factorizable subnetworks?
To find out, we masked out PLM weights to suppress performance on CATH subcategories or secondary structure elements while maintaining performance on other sequences or residues.
To find out, we masked out PLM weights to suppress performance on CATH subcategories or secondary structure elements while maintaining performance on other sequences or residues.
Comments
@avapamini.bsky.social @lcrawford.bsky.social and me!
Paper: https://www.biorxiv.org/content/10.1101/2025.05.29.656902v1
Code: https://github.com/microsoft/plm_subnetworks