Tag
This paper introduces an online variant of assistance games and provides the first provably efficient learning algorithms for both the human and assistant agents, achieving near-optimal regret bounds.