Behavioural vs. Representational Systematicity in End-to-End Models: An Opinionated Survey

TL;DR

This survey critically examines the distinction between behavioural and representational systematicity in end-to-end models, analyzing existing benchmarks and proposing ways to assess the systematicity of internal representations.

Contribution

It clarifies the difference between behavioural and representational systematicity and evaluates how current benchmarks test these aspects in language and vision models.

Findings

Most benchmarks focus on behavioural systematicity.

Representational systematicity is less directly assessed.

The paper suggests methods from mechanistic interpretability for evaluating internal representations.

Abstract

A core aspect of compositionality, systematicity is a desirable property in ML models as it enables strong generalization to novel contexts. This has led to numerous studies proposing benchmarks to assess systematic generalization, as well as models and training regimes designed to enhance it. Many of these efforts are framed as addressing the challenge posed by Fodor and Pylyshyn. However, while they argue for systematicity of representations, existing benchmarks and models primarily focus on the systematicity of behaviour. We emphasize the crucial nature of this distinction. Furthermore, building on Hadley's (1994) taxonomy of systematic generalization, we analyze the extent to which behavioural systematicity is tested by key benchmarks in the literature across language and vision. Finally, we highlight ways of assessing systematicity of representations in ML models as practiced in the field of mechanistic interpretability.