Liam Mertens Phi H

New research highlights limitations in multi-objective RL timescales

This paper introduces a new perspective on multi-objective reinforcement learning (RL) and successor features, arguing that current approaches are insufficient. The authors contend that existing methods fail to account …