Our work on algorithm and data dependent generalization bounds for diffusion models is accepted to NeurIPS 2025!