SEP 9, 2026 · PREPRINT
Meta-LinEXP3: Online-within-Online Learning for Adversarial Linear Contextual Bandits
arXiv
This is a theoretical computer science contribution proposing a novel algorithm with regret bounds and limited experimental validation; it raises and addresses a computational learning problem rather than answering a clinical or established empirical question.
Reported
Per-task regret (known distributi…O(√n)
Leading regret term (unknown dist…O(n^(2/3))