Understanding Domain Learning in Language Models Through Subpopulation Analysis

TL;DR

This paper explores how language models encode different domains by analyzing their internal representations across layers using subpopulation analysis and SVCCA, revealing how model size influences domain information storage.

Contribution

It introduces a novel subpopulation analysis method with SVCCA to study domain encoding in Transformer models, highlighting the effects of model capacity on domain representation.

Findings

Larger models store domain information differently across layers.

Experimental models embed multiple domains similarly to conjoined control models.

The method is validated through qualitative analysis.

Abstract

We investigate how different domains are encoded in modern neural network architectures. We analyze the relationship between natural language domains, model size, and the amount of training data used. The primary analysis tool we develop is based on subpopulation analysis with Singular Vector Canonical Correlation Analysis (SVCCA), which we apply to Transformer-based language models (LMs). We compare the latent representations of such a language model at its different layers from a pair of models: a model trained on multiple domains (an experimental model) and a model trained on a single domain (a control model). Through our method, we find that increasing the model capacity impacts how domain information is stored in upper and lower layers differently. In addition, we show that larger experimental models simultaneously embed domain-specific information as if they were conjoined control models. These findings are confirmed qualitatively, demonstrating the validity of our method.