In-Context Learning for Pure Exploration

Alessio Russo, Ryan Welch, Aldo Pacchiano

We study the problem active sequential hypothesis testing, also known as pure exploration: given a new task, the learner adaptively collects data from the environment to efficientl…