Attention, May I Have Your Decision? Localizing Generative Choices in Diffusion Models
For researchers and practitioners of text-to-image generation, this work provides a method to control implicit biases in diffusion models by targeting specific layers, improving fairness and reducing artifacts.
This work localizes implicit decision-making in text-to-image diffusion models to self-attention layers, and proposes ICM, a steering method that intervenes on these layers to achieve superior debiasing performance with fewer artifacts compared to existing methods.
Text-to-image diffusion models exhibit remarkable generative capabilities, yet their internal operations remain opaque, particularly when handling prompts that are not fully descriptive. In such scenarios, models must make implicit decisions to generate details not explicitly specified in the text. This work investigates the hypothesis that this decision-making process is not diffuse but is computationally localized within the model's architecture. While existing localization techniques focus on prompt-related interventions, we notice that such explicit conditioning may differ from implicit decisions. Therefore, we introduce a probing-based localization technique to identify the layers with the highest attribute separability for concepts. Our findings indicate that the resolution of ambiguous concepts is governed principally by self-attention layers, identifying them as the most effective point for intervention. Based on this discovery, we propose ICM (Implicit Choice-Modification) - a precise steering method that applies targeted interventions to a small subset of layers. Extensive experiments confirm that intervening on these specific self-attention layers yields superior debiasing performance compared to existing state-of-the-art methods, minimizing artifacts common to less precise approaches. The code is available at https://github.com/kzaleskaa/icm.