How do diffusion models work? Select anyone? 1. None of the listed options. 2. They gradually add noise into the original image. 3. simplifies integration of LLMs in projects. 4. Manage memory.

Diffusion models work by gradually adding noise to an image in a controlled manner and then learning to reverse this process to generate new images. Thus, the correct option among the ones you provided is: **2. They gradually add noise into the original image.** In detail, diffusion models start with an original image and iteratively add Gaussian noise over several time steps. During training, the model learns to predict and remove this noise step-by-step, eventually reconstructing the original image from the noisy version. At inference time, the model can generate new images by starting from pure noise and using the learned denoising process to create realistic samples.