Neural organization of sensory stimuli into decision-relevant variables

Determine how the human brain structures sensory stimuli into decision-relevant variables that enable efficient control and adaptation.

Background

The paper motivates its contributions by highlighting a central uncertainty in cognitive neuroscience: despite advances in reinforcement learning and representation learning, the mechanisms by which the brain efficiently organizes multimodal sensory input for planning and control are not well understood.

This unresolved question motivates the authors’ pursuit of an algorithmic framework that leverages active inference principles within distributional RL to better exploit efficient information organization without explicit transition-model learning.