Consciousness-Inspired Spatio-Temporal Abstractions for Better Generalization in Reinforcement Learning

Mingde Zhao, Safa Alver, H. V. Seijen, Romain Laroche, D. Precup +1 more
2/12/2026

No official code found. Best implementation: mila-iqia/Skipper (27 stars, MIT, last commit 2024-09-25).

Abstract

Inspired by human conscious planning, we propose Skipper, a model-based reinforcement learning framework utilizing spatio-temporal abstractions to generalize better in novel situations. It automatically decomposes the given task into smaller, more manageable subtasks, and thus enables sparse decision-making and focused computation on the relevant parts of the environment. The decomposition relies on the extraction of an abstracted proxy problem represented as a directed graph, in which vertices and edges are learned end-to-end from hindsight. Our theoretical analyses provide performance guarantees under appropriate assumptions and establish where our approach is expected to be helpful. Generalization-focused experiments validate Skipper's significant advantage in zero-shot generalization, compared to some existing state-of-the-art hierarchical planning methods.

View on arXivView PDF

Code Implementations(1)

mila-iqia/SkipperCommunity100%
MIT

A PyTorch Implementation of Skipper

271PythonSep 29, 20232 years agoMIT

Similar Papers

Consciousness-Inspired Spatio-Temporal Abstractions for Better Generalization in Reinforcement Learning

100% match1 repo272023

Brain-Inspired Planning for Better Generalization in Reinforcement Learning

91% match2025

A New View on Planning in Online Reinforcement Learning

84% match2024

Reconciling Spatial and Temporal Abstractions for Goal Representation

83% match1 repo72024

Toward Task Generalization via Memory Augmentation in Meta-Reinforcement Learning

83% match2025

Spatiotemporal Forecasting as Planning: A Model-Based Reinforcement Learning Approach with Generative World Models

83% match2025

Cite this paper

@article{zhao2026consciousnessinspired,
  title  = {Consciousness-Inspired Spatio-Temporal Abstractions for Better Generalization in Reinforcement Learning},
  author = {Mingde Zhao and Safa Alver and H. V. Seijen and Romain Laroche and D. Precup and Y. Bengio},
  year   = {2026},
  eprint = {2310.00229},
  archivePrefix = {arXiv},
  url    = {https://arxiv.org/abs/2310.00229},
  journal = {ICLR 2023 2023}
}

Add a badge to your README

Shows whether official code is linked for this paper.

Code of Paper badge
[![Code of Paper](https://codeofpaper.com/badge/2310.00229.svg)](https://codeofpaper.com/paper/2310.00229)

Discussion