AI RESEARCH
Supervised Learning Has a Necessary Geometric Blind Spot: Theory, Consequences, and Minimal Repair
arXiv CS.LG
•
ArXi:2604.21395v1 Announce Type: new We prove that empirical risk minimisation (ERM) imposes a necessary geometric constraint on learned representations: any encoder that minimises supervised loss must retain non-zero Jacobian sensitivity in directions that are label-correlated in