{
  "id": "2309.15842",
  "version": "v1",
  "published": "2023-09-27T17:59:11.000Z",
  "updated": "2023-09-27T17:59:11.000Z",
  "title": "Exploiting the Signal-Leak Bias in Diffusion Models",
  "authors": [
    "Martin Nicolas Everaert",
    "Athanasios Fitsios",
    "Marco Bocchio",
    "Sami Arpa",
    "Sabine Süsstrunk",
    "Radhakrishna Achanta"
  ],
  "categories": [
    "cs.CV",
    "cs.LG"
  ],
  "abstract": "There is a bias in the inference pipeline of most diffusion models. This bias arises from a signal leak whose distribution deviates from the noise distribution, creating a discrepancy between training and inference processes. We demonstrate that this signal-leak bias is particularly significant when models are tuned to a specific style, causing sub-optimal style matching. Recent research tries to avoid the signal leakage during training. We instead show how we can exploit this signal-leak bias in existing diffusion models to allow more control over the generated images. This enables us to generate images with more varied brightness, and images that better match a desired style or color. By modeling the distribution of the signal leak in the spatial frequency and pixel domains, and including a signal leak in the initial latent, we generate images that better match expected results without any additional training.",
  "revisions": [
    {
      "version": "v1",
      "updated": "2023-09-27T17:59:11.000Z"
    }
  ],
  "analyses": {
    "keywords": [
      "signal-leak bias",
      "generate images",
      "better match expected results",
      "sub-optimal style",
      "inference pipeline"
    ],
    "note": {
      "typesetting": "TeX",
      "pages": 0,
      "language": "en",
      "license": "arXiv",
      "status": "editable"
    }
  }
}