Concept Algebra for (Score-Based) Text-Controlled Generative Models
This work addresses the interpretability and control of generative models for researchers and practitioners, but it is incremental as it builds on existing ideas of disentangled representations.
The paper tackles the problem of understanding and manipulating learned representations in text-guided generative models by formalizing concepts as subspaces and developing a method to identify and algebraically manipulate them, demonstrated with Stable Diffusion examples.
This paper concerns the structure of learned representations in text-guided generative models, focusing on score-based models. A key property of such models is that they can compose disparate concepts in a `disentangled' manner. This suggests these models have internal representations that encode concepts in a `disentangled' manner. Here, we focus on the idea that concepts are encoded as subspaces of some representation space. We formalize what this means, show there's a natural choice for the representation, and develop a simple method for identifying the part of the representation corresponding to a given concept. In particular, this allows us to manipulate the concepts expressed by the model through algebraic manipulation of the representation. We demonstrate the idea with examples using Stable Diffusion. Code in https://github.com/zihao12/concept-algebra-code