Emergent Communication in Continuous Worlds: Self-Organisation of Conceptually Grounded Vocabularies at Scale

The paper presents a methodology for autonomous agents to develop a linguistic convention for referring to entities in their environment through decentralized communication. It demonstrates the scalability and robustness of this approach across 37 diverse datasets, showing that the emergent vocabulary can adapt to perceptual variations and environmental changes. This work is significant for AI practitioners as it provides insights into self-organizing systems that can enhance communication strategies in multi-agent environments.