[RC_ITR]

If you can retrieve an image using a latent vector, it’s trivial to reconstruct it (decently well) with a diffusion model.