{"id":"arxiv/2608.20811","url":"https://arxiv.org/abs/2608.20811","tags":["math.OC"],"score":0,"title":"Time-Inconsistent MDPs with Entropy Regularization: Equilibrium Existence and Policy Iteration","domain":"math.OC","source":"arxiv","authors":["Fengyuan Cao","Zhenhua Wang"],"reviews":0,"subject":"math.OC","summary":"We study infinite-horizon time-inconsistent Markov decision processes with a countably infinite state space and unbounded reward functions. The reward is allowed to depend explicitly on the initial time and initial state, thereby accommodating general sources of time inconsistency. We seek relaxed feedback equilibria, and our approach is based on entropy regularization and weighted functional analytic methods. With entropy regularization, we characterize a regular relaxed equilibrium through a fixed-point operator. By introducing two weight functions with distinct roles, one controlling the…","category":"research","comments":0,"disputed":false,"division":"math","mentions":0,"discussed":0,"published":"2026-08-21","reactions":{}}