ALPaCA: Adapting Llama for Pathology Context Analysis to enable slide-level question answering
ALPaCA is a slide-level large vision-language model framework for whole-slide pathology question answering across cancer types and tissue sites. Trained with 35,913 WSIs and 341,051 question-answer pairs, it combines a LongFormer vision-text adaptor, Gaussian mixture model-based prototyping, and Llama3.1 to support whole-slide reasoning and organ- or disease-specific adaptation.
Recommended citation: Gao Z, He K, Su W, Pang X, Machado I P, Jimenez-Linan M, Rous B, Wang C, Li C, McGough W, Gao S, Zhang D, Gong T, Lu M Y, Mahmood F, Feng M, Li C, Crispin-Ortuzar M. ALPaCA: Adapting Llama for Pathology Context Analysis to enable slide-level question answering[J]. Nature Communications, 2026. https://doi.org/10.1038/s41467-026-76372-z.
Download Paper