Adaptive World Models: Learning Behaviors by Latent Imagination Under Non-Stationarity

Abstract

Developing foundational world models is a key research direction for embodied intelligence, with the ability to adapt to non-stationary environments being a crucial criterion. In this work, we introduce a new formalism, Hidden Parameter-POMDP, designed for control with adaptive world models. We demonstrate that this approach enables learning robust behaviors across a variety of non-stationary RL benchmarks. Additionally, this formalism effectively learns task abstractions in an unsupervised manner, resulting in structured, task-aware latent spaces.

PDF NeurIPSW OpenReview Semantic Scholar

Cite

Text

Gospodinov et al. "Adaptive World Models: Learning Behaviors by Latent Imagination Under Non-Stationarity." NeurIPS 2024 Workshops: AFM, 2024.

Markdown

[Gospodinov et al. "Adaptive World Models: Learning Behaviors by Latent Imagination Under Non-Stationarity." NeurIPS 2024 Workshops: AFM, 2024.](https://mlanthology.org/neuripsw/2024/gospodinov2024neuripsw-adaptive/)

BibTeX

@inproceedings{gospodinov2024neuripsw-adaptive,
  title     = {{Adaptive World Models: Learning Behaviors by Latent Imagination Under Non-Stationarity}},
  author    = {Gospodinov, Emiliyan and Shaj, Vaisakh and Becker, Philipp and Geyer, Stefan and Neumann, Gerhard},
  booktitle = {NeurIPS 2024 Workshops: AFM},
  year      = {2024},
  url       = {https://mlanthology.org/neuripsw/2024/gospodinov2024neuripsw-adaptive/}
}