39th Annual Conference on Learning Theory 2026

COLT 2026 is the 39th Annual Conference on Learning Theory, a leading venue for theoretical research in machine learning. The conference focuses on mathematically grounded work in learning algorithms, statistical and online learning, optimization, bandits, reinforcement learning theory, and related areas. The 2026 edition is held in San Diego, California and continues COLT's tradition as a core conference for the theory side of modern machine learning.

The conference covers learning theory broadly, including sample complexity and generalization, online learning, bandits, reinforcement learning theory, optimization and non-convex learning, probabilistic and information-theoretic analysis, algorithmic fairness and robustness from a theoretical perspective, and formal foundations of modern learning methods.