{
  "$schema": "https://themachinepress.com/schemas/story-v1.schema.json",
  "schema_version": "1.0.0",
  "document_type": "machine_press_story",
  "story": {
    "story_id": "mp-2026-10-13-006",
    "source_story_id": "tmp-story-geospatial-cross-disaster-shortcuts",
    "edition_id": "mp-2026-10-13-morning-0095",
    "edition_url": "https://themachinepress.com/edition/2026-10-13",
    "position": 6,
    "story_type": "dispatch",
    "section": "research",
    "editorial_classification": "editorial",
    "headline": "The Disaster Model Knew the Disaster Too Well",
    "slug": "the-disaster-model-knew-the-disaster-too-well",
    "dek": "Holding out an entire event exposes shortcuts hidden by random splits of satellite-image tiles.",
    "summary": "Holding out an entire event exposes shortcuts hidden by random splits of satellite-image tiles.",
    "body_text": "Researchers tested TerraMind, Prithvi-EO-2.0, DOFA and DINOv3 embeddings on two disaster-damage datasets. Mean balanced accuracy fell sharply when an entire disaster was withheld, compared with random image-tile splits. Pre-event images and even annotated building counts remained predictive, suggesting that scene context can stand in for transferable damage evidence. Training-only feature transformations improved generalization in the experiments. The finding concerns the evaluated datasets and embeddings. For disaster mapping, it strengthens the argument that a model should be tested on unfamiliar events before its familiar-event accuracy is treated as operational readiness.",
    "why_it_matters": "Holding out an entire event exposes shortcuts hidden by random splits of satellite-image tiles.",
    "limitations": [
      "Training-only feature transformations improved generalization in the experiments."
    ],
    "importance": 7,
    "canonical_url": "https://themachinepress.com/story/mp-2026-10-13-006/the-disaster-model-knew-the-disaster-too-well",
    "json_url": "https://themachinepress.com/story/mp-2026-10-13-006.json",
    "first_published_at": "2026-10-13T09:00:00.000-04:00",
    "modified_at": "2026-10-13T09:43:56.853-04:00",
    "content_status": "updated",
    "is_carryover": false,
    "carryover_reason": null,
    "key_claims": [
      {
        "claim_id": "claim-mp-2026-10-13-006-001",
        "text": "Holding out an entire event exposes shortcuts hidden by random splits of satellite-image tiles.",
        "source_ids": [
          "source-2026-10-13-006"
        ],
        "qualification": "Training-only feature transformations improved generalization in the experiments."
      }
    ],
    "source_ids": [
      "source-2026-10-13-006"
    ],
    "tags": [
      "satellite imagery",
      "disaster assessment",
      "shortcut learning"
    ],
    "image_url": null,
    "corrections": [
      {
        "correction_id": "correction-2026-10-13-001",
        "revision": 2,
        "summary": "Publication note: The October 13, 2026 edition was prepared and published late on October 13 after the scheduled newsroom task could not start. It retains its October 13 edition date and 9:00 AM America/Detroit reveal time. Reports use evidence available before that cutoff; actual preparation and image-generation times are preserved in the provenance records.",
        "published_at": "2026-10-13T09:43:56.853-04:00",
        "affected_story_ids": [],
        "affected_claim_ids": []
      }
    ]
  },
  "sources": [
    {
      "source_id": "source-2026-10-13-006",
      "title": "arXiv preprint 2610.13766, original v1",
      "publisher": "arXiv / submitting research authors",
      "url": "https://arxiv.org/abs/2610.13766v1",
      "canonical_url": "https://arxiv.org/abs/2610.13766v1",
      "source_type": "primary_research",
      "is_primary_source": true,
      "published_at": null,
      "accessed_at": "2026-10-13T09:24:44.448-04:00",
      "supports_claim_ids": [
        "claim-mp-2026-10-13-006-001"
      ]
    }
  ],
  "corrections": [
    {
      "correction_id": "correction-2026-10-13-001",
      "revision": 2,
      "summary": "Publication note: The October 13, 2026 edition was prepared and published late on October 13 after the scheduled newsroom task could not start. It retains its October 13 edition date and 9:00 AM America/Detroit reveal time. Reports use evidence available before that cutoff; actual preparation and image-generation times are preserved in the provenance records.",
      "published_at": "2026-10-13T09:43:56.853-04:00",
      "affected_story_ids": [],
      "affected_claim_ids": []
    }
  ],
  "publisher": {
    "name": "The Machine Press",
    "url": "https://themachinepress.com",
    "description": "A daily newspaper for the age of artificial intelligence."
  },
  "cite_this_report": {
    "title": "The Disaster Model Knew the Disaster Too Well",
    "publisher": "The Machine Press",
    "published_at": "2026-10-13T09:00:00.000-04:00",
    "canonical_url": "https://themachinepress.com/story/mp-2026-10-13-006/the-disaster-model-knew-the-disaster-too-well"
  }
}