# Evidence of an Emergent "Self" in Continual Robot Learning

arXiv: https://arxiv.org/abs/2603.24350 · HTML: https://mlsanity.com/paper/2603.24350/

## Key facts

- No citing papers in the map yet (data through 14 Oct 2026).
- Builds on 29 of its 53 references that are in the map.
- First posted on arXiv 25 Mar 2026.

## Builds on (most cited first, 29 of 29)

- [Soft Actor-Critic: Off-Policy Maximum Entropy Deep Reinforcement Learning with a Stochastic Actor](https://mlsanity.com/paper/1801.01290/) (2018)
- [Overcoming catastrophic forgetting in neural networks](https://mlsanity.com/paper/1612.00796/) (2016)
- [Similarity of Neural Network Representations Revisited](https://mlsanity.com/paper/1905.00414/) (2019)
- [Understanding deep learning requires rethinking generalization](https://mlsanity.com/paper/1611.03530/) (2016)
- [Progressive Neural Networks](https://mlsanity.com/paper/1606.04671/) (2016)
- [How transferable are features in deep neural networks?](https://mlsanity.com/paper/1411.1792/) (2014)
- [Meta-World: A Benchmark and Evaluation for Multi-Task and Meta Reinforcement Learning](https://mlsanity.com/paper/1910.10897/) (2019)
- [Git Re-Basin: Merging Models modulo Permutation Symmetries](https://mlsanity.com/paper/2209.04836/) (2022)
- [PathNet: Evolution Channels Gradient Descent in Super Neural Networks](https://mlsanity.com/paper/1701.08734/) (2017)
- [Quantifying Generalization in Reinforcement Learning](https://mlsanity.com/paper/1812.02341/) (2018)
- [The Role of Permutation Invariance in Linear Mode Connectivity of Neural Networks](https://mlsanity.com/paper/2110.06296/) (2021)
- [Towards Continual Reinforcement Learning: A Review and Perspectives](https://mlsanity.com/paper/2012.13490/) (2020)
- [Understanding the Role of Individual Units in a Deep Neural Network](https://mlsanity.com/paper/2009.05041/) (2020)
- [A Study on Overfitting in Deep Reinforcement Learning](https://mlsanity.com/paper/1804.06893/) (2018)
- [Explainable Reinforcement Learning: A Survey](https://mlsanity.com/paper/2005.06247/) (2020)
- [Geometry of the Loss Landscape in Overparameterized Neural Networks: Symmetries and Invariances](https://mlsanity.com/paper/2105.12221/) (2021)
- [Continual World: A Robotic Benchmark For Continual Reinforcement Learning](https://mlsanity.com/paper/2105.10919/) (2021)
- [On the importance of single directions for generalization](https://mlsanity.com/paper/1803.06959/) (2018)
- [Sharing Knowledge in Multi-Task Deep Reinforcement Learning](https://mlsanity.com/paper/2401.09561/) (2024)
- [A Survey of Explainable Reinforcement Learning](https://mlsanity.com/paper/2202.08434/) (2022)
- [Learning Shared Representations in Multi-task Reinforcement Learning](https://mlsanity.com/paper/1603.02041/) (2016)
- [Full-Body Visual Self-Modeling of Robot Morphologies](https://mlsanity.com/paper/2111.06389/) (2021)
- [AI Must Embrace Specialization via Superhuman Adaptable Intelligence](https://mlsanity.com/paper/2602.23643/) (2026)
- [Distilling Reinforcement Learning Policies for Interpretable Robot Locomotion: Gradient Boosting Machines and Symbolic Regression](https://mlsanity.com/paper/2403.14328/) (2024)
- [Exploring Human-AI Conceptual Alignment through the Prism of Chess](https://mlsanity.com/paper/2510.26025/) (2025)
- [Continual Learning with Neuron Activation Importance](https://mlsanity.com/paper/2107.12657/) (2021)
- [Modular Boundaries in Recurrent Neural Networks](https://mlsanity.com/paper/2310.20601/) (2023)
- [Robot Body Schema Learning from Full-body Extero/Proprioception Sensors](https://mlsanity.com/paper/2402.18675/) (2024)
- [Robot Metabolism: Towards machines that can grow by consuming other machines](https://mlsanity.com/paper/2411.11192/) (2024)

## Abstract (arXiv)

A key challenge to understanding self-awareness has been a principled way of quantifying whether an intelligent system has a concept of a "self", and if so how to differentiate the "self" from other cognitive structures. We propose that the "self" can be isolated by seeking the invariant portion of cognitive process that changes relatively little compared to more rapidly acquired cognitive skills - because our self is the most persistent aspect of our experiences. We used this principle to analyze the cognitive structure of robots under two conditions: One robot learns a constant task, while a second undergoes continual learning under variable tasks. We find that robots subjected to continual learning develop an invariant subnetwork that is significantly more stable (p < 0.001) compared to the control, and that this subnetwork is also functionally important: preserving it aids adaptation while damaging it impairs performance. We validate this pattern across three different robots spanning locomotion and manipulation.