EpiBench is a sequence-based benchmark featuring 1,609 curated samples to evaluate if LLMs can infer epitope information from antigen and antibody sequences. It measures reasoning capabilities grounded in structural contacts and functional B-cell assays.
HOW THIS AFFECTS YOU
●
researcherYou can use this benchmark to evaluate the biological reasoning limits of protein-centric LLMs.
●
healthThis provides a standardized way to assess AI utility in antibody drug discovery workflows.