Abstract page for arXiv paper 2610.10859: Enabling Preference-driven Unlearning in Few-step Distilled Text-to-Image Diffusion Models