AI RESEARCH
IdEst: Assessing Self-Supervised Learning Representations via Intrinsic Dimension
arXiv CS.LG
•
ArXi:2606.03338v1 Announce Type: new Self-supervised learning (SSL) has emerged as a powerful paradigm for learning meaningful representations from unlabeled data. However, the standard protocol for evaluating these representations, linear probing, is computationally expensive, sensitive to hyperparameters, and provides limited insight into the geometric structure of the representation space.