The interaction is usually modeled as Markov Decision Processes (MDPs). Research on MDPs can be broadly divided into two lines based on the reward generation mechanism. The first line of work [Jaksch et al., 2010, Azar et al., 2013, 2017, He et al., 2021] considers the

We adopt a multi-armed bandit framework for this process, enhancing prompt selection based on the returns from online trajectories.

However, existing frameworks face several interrelated limitations.

The task of sequential memory is considered challenging for models operating under biological constraints (i.e., local synaptic computations) for many reasons, including catastrophic forgetting,