The field of generative models is moving towards increased accountability and trust, with a focus on developing methods for attributing generated content to its source model. This is particularly important in commercial settings where users expect guarantees about the source of the content they receive. Researchers are exploring innovative approaches to capture model-specific signatures and achieve reliable model attribution, even when the model's internal details are inaccessible. Another area of advancement is in the generation of structured images, such as fingerprints and textures, using biologically inspired synchronization dynamics as structured priors. Furthermore, there is a growing interest in applying machine learning to non-traditional domains, such as number theory, to uncover new patterns and insights. Noteworthy papers in this area include: PRISM, which introduces a scalable framework for fingerprinting AI-generated images with high accuracy. Kuramoto Orientation Diffusion Models, which proposes a score-based generative model built on periodic domains to generate structured images. Causal Fingerprints of AI Generative Models, which conceptualizes the causal fingerprint of generative models and proposes a causality-decoupling framework for model attribution. StrCGAN, which introduces a generative framework for stellar image restoration that preserves stellar morphology and generates physically consistent reconstructions.