{
  "$schema": "https://themachinepress.com/schemas/story-v1.schema.json",
  "schema_version": "1.0.0",
  "document_type": "machine_press_story",
  "story": {
    "story_id": "mp-2026-09-01-011",
    "source_story_id": "tmp-story-caer-world-models",
    "edition_id": "mp-2026-09-01-morning-0054",
    "edition_url": "https://themachinepress.com/edition/2026-09-01",
    "position": 11,
    "story_type": "dispatch",
    "section": "robotics",
    "editorial_classification": "editorial",
    "headline": "The World Model Spent More Gradient Where Actions Mattered",
    "slug": "the-world-model-spent-more-gradient-where-actions-mattered",
    "dek": "CAER uses the model's own action-conditioned counterfactual to weight sparse interaction dynamics above static background.",
    "summary": "CAER uses the model's own action-conditioned counterfactual to weight sparse interaction dynamics above static background.",
    "body_text": "Uniform video-reconstruction loss lets abundant background tokens dominate training even when an action changes only a small part of the scene. CAER compares predictions with and without the action, localizes affected tokens online and redistributes a fixed total weight toward them without external labels or preprocessing. The authors report consistent gains in physical consistency, controllability and visual quality across heterogeneous action-conditioned tasks. It is a training paradigm evaluated by its proposing team, not proof of causal understanding outside those tasks.",
    "why_it_matters": "CAER uses the model's own action-conditioned counterfactual to weight sparse interaction dynamics above static background.",
    "limitations": [
      "Uniform video-reconstruction loss lets abundant background tokens dominate training even when an action changes only a small part of the scene.",
      "It is a training paradigm evaluated by its proposing team, not proof of causal understanding outside those tasks."
    ],
    "importance": 8,
    "canonical_url": "https://themachinepress.com/story/mp-2026-09-01-011/the-world-model-spent-more-gradient-where-actions-mattered",
    "json_url": "https://themachinepress.com/story/mp-2026-09-01-011.json",
    "first_published_at": "2026-09-01T09:00:00.000-04:00",
    "modified_at": "2026-09-01T09:00:00.000-04:00",
    "content_status": "new",
    "is_carryover": false,
    "carryover_reason": null,
    "key_claims": [
      {
        "claim_id": "claim-mp-2026-09-01-011-001",
        "text": "CAER uses the model's own action-conditioned counterfactual to weight sparse interaction dynamics above static background.",
        "source_ids": [
          "source-2026-09-01-011"
        ],
        "qualification": "Uniform video-reconstruction loss lets abundant background tokens dominate training even when an action changes only a small part of the scene."
      }
    ],
    "source_ids": [
      "source-2026-09-01-011"
    ],
    "tags": [
      "world models",
      "action conditioning",
      "video generation"
    ],
    "image_url": null,
    "corrections": []
  },
  "sources": [
    {
      "source_id": "source-2026-09-01-011",
      "title": "arXiv preprint 2608.30897",
      "publisher": "arXiv",
      "url": "https://arxiv.org/abs/2608.30897",
      "canonical_url": "https://arxiv.org/abs/2608.30897",
      "source_type": "primary_research",
      "is_primary_source": true,
      "published_at": "2026-08-31T10:49:56.000-04:00",
      "accessed_at": "2026-09-01T08:25:35.000-04:00",
      "supports_claim_ids": [
        "claim-mp-2026-09-01-011-001"
      ]
    }
  ],
  "corrections": [],
  "publisher": {
    "name": "The Machine Press",
    "url": "https://themachinepress.com",
    "description": "A daily newspaper for the age of artificial intelligence."
  },
  "cite_this_report": {
    "title": "The World Model Spent More Gradient Where Actions Mattered",
    "publisher": "The Machine Press",
    "published_at": "2026-09-01T09:00:00.000-04:00",
    "canonical_url": "https://themachinepress.com/story/mp-2026-09-01-011/the-world-model-spent-more-gradient-where-actions-mattered"
  }
}