The 38th Annual Conference on Learning Theory (COLT 2025) will take place June 30th-July 4th, 2025 in Lyon, France. We invite submissions of papers addressing theoretical aspects of machine learning, broadly defined as a subject at the intersection of computer science, statistics and applied mathematics. We strongly support an inclusive view of learning theory, including fundamental theoretical aspects of learnability in various contexts, and theory that sheds light on empirical phenomena. The topics include but are not limited to: Design and analysis of learning algorithms Statistical and computational complexity of learning Optimization methods for learning, including online and stochastic optimization Theory of artificial neural networks, including deep learning Theoretical explanation of empirical phenomena in learning Supervised learning Unsupervised, semi-supervised learning, domain adaptation Learning geometric and topological structures in data, manifold learning Active and interactive learning Reinforcement learning Online learning and decision-making Interactions of learning theory with other mathematical fields High-dimensional and non-parametric statistics Kernel methods Causality Theoretical analysis of probabilistic graphical models Bayesian methods in learning Game theory and learning Learning with system constraints (e.g., privacy, fairness, memory, communication) Learning from complex data (e.g., networks, time series) Learning in neuroscience, social science, economics and other subjects Submissions by authors who are new to COLT are encouraged. While the primary focus of the conference is theoretical, authors are welcome to support their analysis with relevant experimental results. Accepted papers will be presented at the conference. At least one author of each accepted paper should present the work at the conference. Accepted papers will be published electronically in the Proceedings of Machine Learning Research (PMLR). Authors of accepted papers will have the option of opting out of the proceedings in favor of a 1-page extended abstract, which will point to an open access archival version of the full paper reviewed for COLT.