Developmentally motivated emergence of compositional communication via template transfer
This addresses the challenge of enabling agents to develop structured communication without architectural biases, with potential applications in AI and robotics, though it appears incremental compared to existing methods.
The paper tackles the problem of achieving emergent compositional communication in multi-agent systems by proposing a template transfer training regime, resulting in demonstrated emergence using metrics like topographical similarity and zero-shot generalization.
This paper explores a novel approach to achieving emergent compositional communication in multi-agent systems. We propose a training regime implementing template transfer, the idea of carrying over learned biases across contexts. In our method, a sender-receiver pair is first trained with disentangled loss functions and then the receiver is transferred to train a new sender with a standard loss. Unlike other methods (e.g. the obverter algorithm), our approach does not require imposing inductive biases on the architecture of the agents. We experimentally show the emergence of compositional communication using topographical similarity, zero-shot generalization and context independence as evaluation metrics. The presented approach is connected to an important line of work in semiotics and developmental psycholinguistics: it supports a conjecture that compositional communication is scaffolded on simpler communication protocols.