{
  "$schema": "https://themachinepress.com/schemas/story-v1.schema.json",
  "schema_version": "1.0.0",
  "document_type": "machine_press_story",
  "story": {
    "story_id": "mp-2026-10-02-017",
    "source_story_id": "tmp-sidebar-dmad-adversarial-distillation",
    "edition_id": "mp-2026-10-02-morning-0085",
    "edition_url": "https://themachinepress.com/edition/2026-10-02",
    "position": 19,
    "story_type": "ticker",
    "section": "media-creative-tools",
    "editorial_classification": "editorial",
    "headline": "A Discriminator Replaced the Student's Extra Diffusion Model",
    "slug": "a-discriminator-replaced-the-student-s-extra-diffusion-model",
    "dek": "DMAD matched distributions through log-density ratios and reached one-step ImageNet FID 1.04.",
    "summary": "DMAD matched distributions through log-density ratios and reached one-step ImageNet FID 1.04.",
    "body_text": "DMAD learns the density-ratio signal for few-step visual distillation with two discriminator heads instead of maintaining an auxiliary student-score model. The authors report leading compared image, video and audio-video results, including one-step ImageNet-64 FID 1.04.",
    "why_it_matters": "DMAD matched distributions through log-density ratios and reached one-step ImageNet FID 1.04.",
    "limitations": [],
    "importance": 8,
    "canonical_url": "https://themachinepress.com/story/mp-2026-10-02-017/a-discriminator-replaced-the-student-s-extra-diffusion-model",
    "json_url": "https://themachinepress.com/story/mp-2026-10-02-017.json",
    "first_published_at": "2026-10-02T09:00:00.000-04:00",
    "modified_at": "2026-10-02T09:00:00.000-04:00",
    "content_status": "new",
    "is_carryover": false,
    "carryover_reason": null,
    "key_claims": [
      {
        "claim_id": "claim-mp-2026-10-02-017-001",
        "text": "DMAD matched distributions through log-density ratios and reached one-step ImageNet FID 1.04.",
        "source_ids": [
          "source-2026-10-02-019"
        ],
        "qualification": null
      }
    ],
    "source_ids": [
      "source-2026-10-02-019"
    ],
    "tags": [
      "diffusion",
      "distillation",
      "visual generation"
    ],
    "image_url": null,
    "corrections": []
  },
  "sources": [
    {
      "source_id": "source-2026-10-02-019",
      "title": "arXiv preprint 2610.02188",
      "publisher": "arXiv",
      "url": "https://arxiv.org/abs/2610.02188",
      "canonical_url": "https://arxiv.org/abs/2610.02188",
      "source_type": "primary_research",
      "is_primary_source": true,
      "published_at": "2026-10-01T13:59:01.000-04:00",
      "accessed_at": "2026-10-02T08:44:06.000-04:00",
      "supports_claim_ids": [
        "claim-mp-2026-10-02-017-001"
      ]
    }
  ],
  "corrections": [],
  "publisher": {
    "name": "The Machine Press",
    "url": "https://themachinepress.com",
    "description": "A daily newspaper for the age of artificial intelligence."
  },
  "cite_this_report": {
    "title": "A Discriminator Replaced the Student's Extra Diffusion Model",
    "publisher": "The Machine Press",
    "published_at": "2026-10-02T09:00:00.000-04:00",
    "canonical_url": "https://themachinepress.com/story/mp-2026-10-02-017/a-discriminator-replaced-the-student-s-extra-diffusion-model"
  }
}