The universal weight subspace hypothesis
Computer Science > Machine Learning arXiv:2512.05117 (cs) [Submitted on 4 Dec 2025 (v1), last revised 6 Dec 2025 (this version, v2)] Title:The Universal Weight Subspace Hypothesis Authors:Prakhar Kaushik, Shravan Chaudhari, Ankit Vaidya, Rama Chellappa, Alan Yuille View a PDF of the paper titled The Universal Weight Subspace Hypothesis, by Prakhar Kaushik and 4 other authors View PDF HTML (experimental) Abstract:We show that deep neural networks trained across diverse tasks exhibit remarkably similar low-dimensional parametric subspaces. We provide the first large-scale empirical evidence that demonstrates that neural networks systematically converge to shared spectral subspaces regardless of initialization, task, or domain. Through mode-wise spectral analysis of over 1100…
