AI RESEARCH
The Loss Is Not Enough: Sampling Conditions and Inductive Bias in Contrastive Representation Learning
arXiv CS.AI
•
ArXi:2606.04280v1 Announce Type: cross Contrastive learning has become a leading paradigm for self-supervised representation learning, yet the conditions under which it recovers meaningful latent geometry remain incompletely understood. We develop a measure-theoretic framework formalizing the diversity condition, a requirement on positive-pair sampling that is necessary for isometric latent recovery.