AI RESEARCH
Does Your Neural Network Extrapolate? Feature Engineering as Identifiability Bias for OOD Generalization
arXiv CS.AI
•
ArXi:2605.07483v1 Announce Type: cross Successful deep neural networks discover salient features of data. We show when and why they fail to window. This requires decoupling feature learning from data-generating-process (DGP) identifiability. From a single