High-Fidelity Reconstructions of Strong Lenses in the Data-Driven Generative Modeling Era

We achieve state-of-the-art reconstructions of strong gravitational lensing systems from the Sloan Lens ACS (SLACS) survey by leveraging score-based diffusion models as high-dimensional priors over major components of the lensing system: the background source, foreground lens light, and point-spread function (PSF). Our approach produces high-resolution models that substantially reduce residuals compared to previous lens modeling attempts. To our knowledge, this is the first application of data-driven generative priors to real strong lensing observations, establishing a new benchmark for precision lens modeling in preparation for upcoming large-scale imaging surveys.