We propose unified latent diffusion framework for simultaneous music generation, source imputation, and query-driven arbitrary source extraction.

A training-free, generative approach that infers object removal order by exploiting statistical co-occurrence and asymmetry priors learned by generative models.

We designed and tested an autoencoder model for stitching together multi-area multi-animal neuronal recording datasets and inpainting neural dynamics from unobserved brain areas in each experiment.