The Latent Color Subspace: Emergent Order in High-Dimensional Chaos
2026-03-12 • Machine Learning
Machine LearningArtificial IntelligenceComputer Vision and Pattern Recognition
AI summaryⓘ
The authors studied how color is represented inside a type of AI model called a Variational Autoencoder used in text-to-image generation. They found that the model’s hidden space organizes colors in a way that matches Hue, Saturation, and Lightness, which are common ways to describe colors. Using this insight, they created a simple method to predict and control colors in generated images without extra training. Their approach works by directly manipulating the model’s internal representations.
Text-to-image generationVariational AutoencoderLatent spaceHueSaturationLightnessLatent Color SubspaceFLUX modelLatent-space manipulation
Authors
Mateusz Pach, Jessica Bader, Quentin Bouniot, Serge Belongie, Zeynep Akata
Abstract
Text-to-image generation models have advanced rapidly, yet achieving fine-grained control over generated images remains difficult, largely due to limited understanding of how semantic information is encoded. We develop an interpretation of the color representation in the Variational Autoencoder latent space of FLUX.1 [Dev], revealing a structure reflecting Hue, Saturation, and Lightness. We verify our Latent Color Subspace (LCS) interpretation by demonstrating that it can both predict and explicitly control color, introducing a fully training-free method in FLUX based solely on closed-form latent-space manipulation. Code is available at https://github.com/ExplainableML/LCS.