Entropic Confinement and Mode Connectivity in Overparameterized Neural Networks

ArXi:2512.06297v2 Announce Type: replace-cross Modern neural networks exhibit a striking property: basins of attraction in the loss landscape are often connected by low-loss paths, yet optimization dynamics generally remain confined to a single convex basin and rarely explore intermediate points. We resolve this paradox by identifying entropic barriers arising from the interplay between curvature variations along these paths and noise in optimization dynamics.