{"id":"arxiv/2607.07538","url":"https://arxiv.org/abs/2607.07538","tags":["cs.LG","math.AP","stat.ML"],"score":0,"title":"Avoiding unsafe sets when training with Langevin Dynamics","domain":"cs.LG","source":"arxiv","authors":["Adam M. Oberman"],"reviews":0,"subject":"cs.LG","summary":"Training a model with noisy gradient descent can be idealized as overdamped Langevin dynamics, and a natural safety question is to bound the probability $ν_t(\\mathcal{A}_H) = \\mathbb{P}(Q_t \\in \\mathcal{A}_H)$ that the trajectory lies in a designated failure region $\\mathcal{A}_H$. We study this for a smooth, strongly convex loss in $d$ dimensions, with $\\mathcal{A}_H$ separated from the minimizer by an energy gap. At the end of training, the equilibrium mass $π(\\mathcal{A}_H)$ is exponentially small in $d$, with a complementary energy-barrier rate when the noise is small. Along the…","category":"research","comments":0,"disputed":false,"division":"cs","mentions":0,"discussed":0,"published":"2026-07-08","reactions":{}}