All Roads Lead to Rome? Exploring Representational Similarities Between Latent Spaces of Generative Image Models

Charumathi Badrinath, Usha Bhalla, Alex Oesterling, Suraj Srinivas, Himabindu Lakkaraju

Keywords: Generative Models, Latent Space, Representations

TL;DR: We compare the latent spaces of several generative image models and find that they contain similar representations.

Abstract: Do different generative image models secretly learn similar underlying representations? We investigate this by measuring the latent space similarity of four different models: VAEs, GANs, Normalizing Flows (NFs), and Diffusion Models (DMs). Our methodology involves training linear maps between frozen latent spaces to "stitch" arbitrary pairs of encoders and decoders and measuring output-based and probe-based metrics on the resulting "stitched'' models. Our main findings are that linear maps between latent spaces of performant models preserve most visual information even when latent sizes differ; for CelebA models, gender is the most similarly represented probe-able attribute. Finally we show on an NF that latent space representations converge early in training.

Student Paper: Yes

Submission Number: 39