Catalytic Role Of Noise And Necessity Of Inductive Biases In The Emergence Of Compositional Communication
This work addresses the fundamental challenge of enabling AI agents to develop structured communication, which is incremental as it builds on existing signaling game frameworks.
The paper tackles the problem of how compositional communication emerges in signaling games, showing theoretically that both inductive biases and noisy channels are necessary, and experimentally confirming that specific noise levels promote compositionality, with results reported using metrics like topographical similarity and conflict count.
Communication is compositional if complex signals can be represented as a combination of simpler subparts. In this paper, we theoretically show that inductive biases on both the training framework and the data are needed to develop a compositional communication. Moreover, we prove that compositionality spontaneously arises in the signaling games, where agents communicate over a noisy channel. We experimentally confirm that a range of noise levels, which depends on the model and the data, indeed promotes compositionality. Finally, we provide a comprehensive study of this dependence and report results in terms of recently studied compositionality metrics: topographical similarity, conflict count, and context independence.