{
  "$schema": "https://themachinepress.com/schemas/story-v1.schema.json",
  "schema_version": "1.0.0",
  "document_type": "machine_press_story",
  "story": {
    "story_id": "mp-2026-10-09-014",
    "source_story_id": "tmp-story-spacecast-predictive-spatial-reasoning",
    "edition_id": "mp-2026-10-09-morning-0091",
    "edition_url": "https://themachinepress.com/edition/2026-10-09",
    "position": 14,
    "story_type": "dispatch",
    "section": "benchmarks-evals",
    "editorial_classification": "editorial",
    "headline": "Seeing the Scene Was Easier Than Predicting Its Next State",
    "slug": "seeing-the-scene-was-easier-than-predicting-its-next-state",
    "dek": "SpaceCast-Bench tests what happens after an intervention, with the strongest evaluated model at 58% versus 87.2% for humans.",
    "summary": "SpaceCast-Bench tests what happens after an intervention, with the strongest evaluated model at 58% versus 87.2% for humans.",
    "body_text": "SpaceCast-Bench moves beyond reading visible spatial relations to an observe-transform-infer sequence. Its 3,862 questions from 182 real scenes span sixteen task types covering perception, local prediction and global prediction. Among twenty-one tested models, the strongest reached 58%, against 87.2% human performance; specialized spatial models remained near chance. Controlled analyses found bridge views useful for combining observations and explicit 3D evidence more consistently helpful than generated outcome images or video. Fine-tuning Qwen3-VL-4B on programmatically generated data raised its benchmark score from 34% to 65.7%, alongside gains across six outside benchmarks. The results separate naming a present relation from updating a scene that can no longer be directly seen.",
    "why_it_matters": "SpaceCast-Bench tests what happens after an intervention, with the strongest evaluated model at 58% versus 87.2% for humans.",
    "limitations": [],
    "importance": 8,
    "canonical_url": "https://themachinepress.com/story/mp-2026-10-09-014/seeing-the-scene-was-easier-than-predicting-its-next-state",
    "json_url": "https://themachinepress.com/story/mp-2026-10-09-014.json",
    "first_published_at": "2026-10-09T09:00:00.000-04:00",
    "modified_at": "2026-10-10T03:43:44.000-04:00",
    "content_status": "new",
    "is_carryover": false,
    "carryover_reason": null,
    "key_claims": [
      {
        "claim_id": "claim-mp-2026-10-09-014-001",
        "text": "SpaceCast-Bench tests what happens after an intervention, with the strongest evaluated model at 58% versus 87.2% for humans.",
        "source_ids": [
          "source-2026-10-09-014"
        ],
        "qualification": null
      }
    ],
    "source_ids": [
      "source-2026-10-09-014"
    ],
    "tags": [
      "spatial reasoning",
      "scene prediction",
      "3D evidence"
    ],
    "image_url": null,
    "corrections": []
  },
  "sources": [
    {
      "source_id": "source-2026-10-09-014",
      "title": "arXiv preprint 2610.12402, original v1",
      "publisher": "arXiv / submitting research authors",
      "url": "https://arxiv.org/abs/2610.12402v1",
      "canonical_url": "https://arxiv.org/abs/2610.12402v1",
      "source_type": "primary_research",
      "is_primary_source": true,
      "published_at": "2026-10-08T13:44:17.000-04:00",
      "accessed_at": "2026-10-10T03:27:42.309-04:00",
      "supports_claim_ids": [
        "claim-mp-2026-10-09-014-001"
      ]
    }
  ],
  "corrections": [],
  "publisher": {
    "name": "The Machine Press",
    "url": "https://themachinepress.com",
    "description": "A daily newspaper for the age of artificial intelligence."
  },
  "cite_this_report": {
    "title": "Seeing the Scene Was Easier Than Predicting Its Next State",
    "publisher": "The Machine Press",
    "published_at": "2026-10-09T09:00:00.000-04:00",
    "canonical_url": "https://themachinepress.com/story/mp-2026-10-09-014/seeing-the-scene-was-easier-than-predicting-its-next-state"
  }
}