# Learning Awareness Models

arXiv: https://arxiv.org/abs/1804.06318 · HTML: https://mlsanity.com/paper/1804.06318/

## Key facts

- Cited by 13 papers in the map (data through 14 Oct 2026).
- First posted on arXiv 17 Apr 2018.

## Built on by (most cited first, 13 of 13)

- [Mind's Eye: Grounded Language Model Reasoning through Simulation](https://mlsanity.com/paper/2210.05359/) (2022)
- [Visuo-Tactile Transformers for Manipulation](https://mlsanity.com/paper/2210.00121/) (2022)
- [Model-based Meta Reinforcement Learning using Graph Structured Surrogate Models](https://mlsanity.com/paper/2102.08291/) (2021)
- [Temporally Abstract Partial Models](https://mlsanity.com/paper/2108.03213/) (2021)
- [A Unifying Framework for Action-Conditional Self-Predictive Reinforcement Learning](https://mlsanity.com/paper/2406.02035/) (2024)
- [Mental Modeling of Reinforcement Learning Agents by Language Models](https://mlsanity.com/paper/2406.18505/) (2024)
- [GrASP: Gradient-Based Affordance Selection for Planning](https://mlsanity.com/paper/2202.04772/) (2022)
- [On the Origins of Self-Modeling](https://mlsanity.com/paper/2209.02010/) (2022)
- [Balancing Value Underestimation and Overestimation with Realistic Actor-Critic](https://mlsanity.com/paper/2110.09712/) (2021)
- [High-Degrees-of-Freedom Dynamic Neural Fields for Robot Self-Modeling and Motion Planning](https://mlsanity.com/paper/2310.03624/) (2023)
- [Double projection for reconstructing dynamical systems: between stochastic and deterministic regimes](https://mlsanity.com/paper/2510.01089/) (2025)
- [Future Predictive Success-or-Failure Classification for Long-Horizon Robotic Tasks](https://mlsanity.com/paper/2404.03415/) (2024)
- [Categorical Policies: Multimodal Policy Learning and Exploration in Continuous Control](https://mlsanity.com/paper/2508.13922/) (2025)

## Abstract (arXiv)

We consider the setting of an agent with a fixed body interacting with an unknown and uncertain external world. We show that models trained to predict proprioceptive information about the agent's body come to represent objects in the external world. In spite of being trained with only internally available signals, these dynamic body models come to represent external objects through the necessity of predicting their effects on the agent's own body. That is, the model learns holistic persistent representations of objects in the world, even though the only training signals are body signals. Our dynamics model is able to successfully predict distributions over 132 sensor readings over 100 steps into the future and we demonstrate that even when the body is no longer in contact with an object, the latent variables of the dynamics model continue to represent its shape. We show that active data collection by maximizing the entropy of predictions about the body---touch sensors, proprioception and vestibular information---leads to learning of dynamic models that show superior performance when used for control. We also collect data from a real robotic hand and show that the same models can be used to answer questions about properties of objects in the real world. Videos with qualitative results of our models are available at https://goo.gl/mZuqAV.