Uncovering Conceptual Blindspots in Generative Image Models Using Sparse Autoencoders

TL;DR

This paper introduces a systematic method using sparse autoencoders to identify and analyze conceptual blindspots in generative image models, revealing specific missing or exaggerated concepts and memorization artifacts.

Contribution

It presents the largest archetypal SAE trained on DINOv2 features for detailed analysis of conceptual disparities in popular generative models.

Findings

Identified suppressed blindspots like bird feeders and DVD discs.

Detected exaggerated blindspots such as wood textures and palm trees.

Isolated memorization artifacts reproducing training data templates.

Abstract

Despite their impressive performance, generative image models trained on large-scale datasets frequently fail to produce images with seemingly simple concepts -- e.g., human hands or objects appearing in groups of four -- that are reasonably expected to appear in the training data. These failure modes have largely been documented anecdotally, leaving open the question of whether they reflect idiosyncratic anomalies or more structural limitations of these models. To address this, we introduce a systematic approach for identifying and characterizing "conceptual blindspots" -- concepts present in the training data but absent or misrepresented in a model's generations. Our method leverages sparse autoencoders (SAEs) to extract interpretable concept embeddings, enabling a quantitative comparison of concept prevalence between real and generated images. We train an archetypal SAE (RA-SAE) on DINOv2 features with 32,000 concepts -- the largest such SAE to date -- enabling fine-grained analysis of conceptual disparities. Applied to four popular generative models (Stable Diffusion 1.5/2.1, PixArt, and Kandinsky), our approach reveals specific suppressed blindspots (e.g., bird feeders, DVD discs, and whitespaces on documents) and exaggerated blindspots (e.g., wood background texture and palm trees). At the individual datapoint level, we further isolate memorization artifacts -- instances where models reproduce highly specific visual templates seen during training. Overall, we propose a theoretically grounded framework for systematically identifying conceptual blindspots in generative models by assessing their conceptual fidelity with respect to the underlying data-generating process.