For each CF and AHo position in a given dataset, we determine whether the predicted mode (i.e. highest-frequency amino acid) from our best model is the same as the actual mode. Results are aggregated
Datasets for Evaluating Expert Specialization in Mixture-of-Experts Antibody Language Models, published in the ICLR 2026 FM4Science Workshop. Files: classification-data.tar.gz: Two classification
We longitudinally sequenced the repertoires of six South African donors 1) prior to infection and through 2) acute and 3) chronic HIV-1 infection and quantified how the antibody repertoire changed ove
adetermined using consensus results from IMGT/V-QUEST and IgBlast.bThe top D gene matches determined by both IMGT/V-QUEST and IgBlast are included.If there were multiple matches, all of the top matche
