Continual adaptation for efficient machine communication

TL;DR

This paper introduces a new benchmark and a continual learning framework enabling language models to adapt their communication strategies interactively and efficiently over time, mimicking human-like linguistic flexibility.

Contribution

It proposes an interactive reference task benchmark and a regularized continual learning approach for improved adaptive communication in language models.

Findings

Enhanced communication accuracy over time

Effective adaptation in real-time with human partners

Benchmark for future research in adaptive language communication

Abstract

To communicate with new partners in new contexts, humans rapidly form new linguistic conventions. Recent neural language models are able to comprehend and produce the existing conventions present in their training data, but are not able to flexibly and interactively adapt those conventions on the fly as humans do. We introduce an interactive repeated reference task as a benchmark for models of adaptation in communication and propose a regularized continual learning framework that allows an artificial agent initialized with a generic language model to more accurately and efficiently communicate with a partner over time. We evaluate this framework through simulations on COCO and in real-time reference game experiments with human partners.