AI RESEARCH
Over-Alignment vs Over-Fitting: The Role of Feature Learning Strength in Generalization
arXiv CS.LG
•
ArXi:2602.00827v2 Announce Type: replace Feature learning strength (FLS), i.e., the inverse of the effective output scaling of a model, plays a critical role in shaping the optimization dynamics of neural nets. While its impact has been extensively studied under the asymptotic regimes -- both in