Reinformed Dreamer: An Asymmetric World Model Trained with Latent-Guided Privileged Information

Forum topic · 小凯 · 2026-07-30

Summary

Reinformed Dreamer (arXiv:2607.26040) is a model-based reinforcement learning algorithm that improves on Dreamer by exploiting privileged information during training. Building on asymmetric reinforcement learning, where agents receive extra supervision beyond rewards, the authors first identify a limitation in how the prior asymmetric algorithm Informed Dreamer learns representations of privileged information. They then propose a novel asymmetric representation learning objective based on latent guidance, yielding Reinformed Dreamer. The method targets both observation representations and privileged information representations, and is applicable under partial observability (extra state information) as well as full observability (finer-grained state information). Experiments across multiple benchmarks show more consistent improvements over Dreamer compared to previous asymmetric approaches. Authors: Gaspard Lambrechts, Adrien Bolland, Daniel Ebi, Damien Ernst; published July 2026.

Paper Overview

Field: Machine Learning Authors: Gaspard Lambrechts, Adrien Bolland, Daniel Ebi, Damien Ernst Published: 2026-07-28 arXiv: 2607.26040

Abstract

Much like humans benefit from guidance while learning, reinforcement learning algorithms may benefit from additional supervision beyond rewards. Leveraging additional information during training to learn better representations and behaviors has been the focus of asymmetric reinforcement learning. This learning paradigm has proven effective under partial observability when additional state information is available, but also under full observability when more refined state information is available.

Focusing on model-based reinforcement learning, the authors study the effect of asymmetric learning on observation representations and on privileged information representations. First, they identify a limitation in the privileged information representations learned by an asymmetric model-based algorithm known as Informed Dreamer. Then, they propose a novel asymmetric representation learning objective using latent guidance, resulting in a new algorithm called Reinformed Dreamer.

Key Contributions

Identifies a limitation of Informed Dreamer in learning representations of privileged information.
Introduces a latent-guided asymmetric representation learning objective.
Proposes Reinforced Dreamer's successor, Reinformed Dreamer, a new asymmetric world-model-based RL algorithm.

Results

Experiments across multiple benchmarks demonstrate that Reinformed Dreamer achieves more consistent improvements over Dreamer than prior asymmetric methods.

---

*Source: forum post, auto-collected 2026-07-30.*