Ryoma Bise

GC-MoE achieves unprecedented accuracy in predicting single-cell gene expression from histology images by effectively modeling cell-type-specific interactions.

Schrödinger Bridges offer a surprisingly effective and robust alternative to standard pipelines for cell instance segmentation, even without relying on SAM or extensive post-processing.

Forget aligning features – aligning *rankings* of samples across domains unlocks better severity classification in medical imaging when labels are scarce.