Reproduce “An Active Inference Model of Covert and Overt Visual Attention” on OpenResearch
Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.
Universal app for Apple silicon and Intel. Requires macOS 11 or later.
About this paper
Tin Mišić, Karlo Koledić, Fabio Bonsignorio, Ivan Petrović, Ivan Marković
Researchers from the University of Zagreb developed an active inference model that effectively integrates covert and overt visual attention, processing raw 2D visual data to reproduce known attentional phenomena. The model demonstrates faster reaction times for valid cues, distinct dynamics for exogenous versus endogenous attention, and emergent inhibition of return, showcasing its applicability to robotics and cognitive modeling.
Key insights
- The model provides a unified framework for covert, overt, endogenous, and exogenous attention, driven by dynamic optimization of sensory precisions to minimize variational free-energy.
- Active inference can mechanistically explain a range of attentional phenomena, including the Posner cueing effect, eccentricity effect, and emergent inhibition of return behavior without explicit modeling.
- Direct processing of raw visual data within active inference offers a principled approach for developing computationally efficient and intelligent autonomous robots, bridging theoretical neuroscience with robotic applications.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.