Thompson Sampling Is 2-Competitive for Mistakes

arXiv:2607.12389 · stat.ML, cs.LG · Submitted 2026-07-14 · Read on arXiv

Mark Sellke, Gregory Valiant

stat.ML, cs.LG

Submitted: 2026-07-14

Comments: 10 pages

License: http://arxiv.org/licenses/nonexclusive-distrib/1.0/

Terminology

• Regret — 7× in this paper · explained in 4 episodes
• Admissibility — 1× in this paper · explained in 2 episodes
• Contextual Bandit — 1× in this paper · explained in 2 episodes
• Contextual Bandits — 1× in this paper · explained in 3 episodes
• Exchangeability — 1× in this paper · explained in 2 episodes
• Regret Bound — 1× in this paper · explained in 2 episodes
• Regret Bounds — 1× in this paper · explained in 2 episodes
• Sample Complexity — 1× in this paper · explained in 20 episodes

Sources

• Approximation Algorithms for Bayesian Multi-Armed Bandit Problems
• On Frequentist Regret of Linear Thompson Sampling

Related papers

• Behavior of prediction performance metrics with rare events
• Optimal Estimation of Generic Dynamics by Path-Dependent Neural Jump ODEs
• A Posterior-Dynamics Framework for Imaging Inverse Problems with Pretrained Diffusion Priors
• Scalable Posterior Uncertainty for Flexible Density-Based Clustering
• A Transport Geometry of Benign Overfitting
• Efficient quadratic entropy with distance sketches