Compositional Understanding in Signaling Games

2025-07-21

AI Summary

In standard signaling games, receivers struggle with compositional semantic interpretation: the loss of any signal component causes complete semantic breakdown. To address this, we propose two novel receiver architectures—the minimalist receiver, which learns atomic signal components in isolation, and the universal receiver, which integrates all components—within an evolutionary learning framework for signaling games. Our core contribution is the decoupling of signal component learning, enabling modular and composable semantic representations. Experimental results demonstrate significantly improved robustness: when partial information is missing, the semantics of remaining components are preserved and accurately decoded. This work achieves, for the first time in evolutionary signaling games, genuine compositional understanding—where meaning emerges systematically from structured combinations of learned primitives, rather than holistic, unstructured associations.

Technology Category

Search and Optimization: Learning to SearchGame Theory and Economic Paradigms: Adversarial LearningMachine Learning: Evolutionary Learning

Application Category

Semantics and Knowledge: Methods to enhance, augment, integrate or synergize semantic models such as knowledge graphs and LLMsSearch and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for rankingEconomics, Online Markets and Human Computation: Social networks and social learning

Abstract

Receivers in standard signaling game models struggle with learning compositional information. Even when the signalers send compositional messages, the receivers do not interpret them compositionally. When information from one message component is lost or forgotten, the information from other components is also erased. In this paper I construct signaling game models in which genuine compositional understanding evolves. I present two new models: a minimalist receiver who only learns from the atomic messages of a signal, and a generalist receiver who learns from all of the available information. These models are in many ways simpler than previous alternatives, and allow the receivers to learn from the atomic components of messages.

Problem

• Receivers struggle with learning compositional signals
• Lost message components erase related information
• New models enable compositional understanding evolution

Innovation

• Minimalist receiver learns atomic messages
• Generalist receiver utilizes all information
• Simpler models enable compositional understanding