Id: arxiv/2602.05259 Url: https://arxiv.org/abs/2602.05259 Tags: math.ST, stat.ML Score: 0 Title: An Asymptotic Law of the Iterated Logarithm for $\mathrm{KL}_{\inf}$ Domain: math.ST Source: arxiv Authors: Ashwin Ram, Aaditya Ramdas Reviews: 0 Subject: math.ST Category: research Comments: 0 Disputed: false Division: math Mentions: 0 Discussed: 0 Published: 2026-02-05 Page: https://delphi.science/r/arxiv/2602.05259 The population $\mathrm{KL}_{\inf}$ is a fundamental quantity that appears in lower bounds for (asymptotically) optimal regret of pure-exploration stochastic bandit algorithms, and optimal stopping time of sequential tests. Motivated by this, an empirical $\mathrm{KL}_{\inf}$ statistic is frequently used in the design of (asymptotically) optimal bandit algorithms and sequential tests. While nonasymptotic concentration bounds for the empirical $\mathrm{KL}_{\inf}$ have been developed, their optimality in terms of constants and rates is questionable, and their generality is limited (usually to…