GPT and Prejudice: A Sparse Approach to Understanding Learned Representations in Large Language Models

TL;DR

It is argued that the LLM+SAEs pipeline offers a scalable framework for auditing how cultural assumptions from the data are embedded in model representations.

Abstract

Large Language Models (LLMs) are trained on massive, unstructured corpora, making it unclear which social patterns and biases they absorb and later reproduce. Existing evaluations typically examine outputs or activations, but rarely connect them back to the pre-training data. We introduce a pipeline that couples LLMs with sparse autoencoders (SAEs) to trace how different themes are encoded during training. As a controlled case study, we trained a GPT-style model on 37 nineteenth-century novels by ten female authors, a corpus centered on themes such as gender, marriage, class, and morality. By applying SAEs across layers and probing with eleven social and moral categories, we mapped sparse features to human-interpretable concepts. The analysis revealed stable thematic backbones (most prominently around gender and kinship) and showed how associations expand and entangle with depth. More broadly, we argue that the LLM+SAEs pipeline offers a scalable framework for auditing how cultural assumptions from the data are embedded in model representations.