BioinformaticsbioRxivPreprint
✦ In-depth digest
BERTology Meets Biology: Interpreting Attention in Protein Language Models
Jesse Vig, Ali Madani, Lav R. Varshney…·Nazneen Fatema Rajani
One attention head of a protein BERT model tracks 3D contacts (28% vs 1.3% background); other heads target binding sites, and deeper layers attend to higher-level properties.
338 citations