AI RESEARCH
Beyond the Failures: Rethinking Foundation Models in Pathology
arXiv CS.CV
•
ArXi:2510.23807v5 Announce Type: replace-cross Despite their successes in vision and language, foundation models have stumbled in pathology, revealing low accuracy, instability, and heavy computational demands. These shortcomings stem not from tuning problems but from deeper conceptual mismatches: dense embeddings cannot represent the combinatorial richness of tissue, and current architectures inherit flaws in self-supervision, patch design, and noise-fragile pre