Sharper Photos, Faster Scans: New AI Fixes What's Blurry

Researchers Darshan Thaker, Lachlan Ewen MacDonald, and René Vidal propose ANaLOG, a framework for solving inverse problems with pretrained latent diffusion models. Native-latent guidance is the paradigm at play: it replaces repeated evaluations of the image-space forward model, each requiring a decoder pass, with efficient guidance computed using a learned latent-space surrogate. But existing methods apply guidance uniformly across latent dimensions, ignoring that measurements are informative only along certain directions and that the reliability of model predictions varies across inputs and timesteps. ANaLOG instead models uncertainty by learning an anisotropic, input- and time-dependent covariance that is integrated into the guidance mechanism to emphasize reliable directions and downweight uncertain ones. The authors theoretically analyze the framework in a linear model setting and prove that anisotropic, uncertainty-aware weighting is necessary for correct sampling, whereas isotropic guidance induces sampling errors. Experiments across five challenging inverse problems show that ANaLOG improves perceptual reconstruction quality over existing methods while preserving efficiency.