Embodied Mortal Agent Learning via Homeostatic Reinforcement Learning

Researchers developed an Embodied Mortal Agent (EMA) that learns linguistic affordances through mortality-grounded reinforcement learning. The model uses homeostatic regulation to link linguistic interaction to an agent's ability to sustain its own existence through social experience.