{
  "schemaVersion": "rnn.event.v1",
  "id": "event-single-arxiv-query-cat-cs-ro-468314c4-veiihza84eblqnighx",
  "slug": "point-tracking-improves-world-action-models",
  "title": "Point Tracking Improves World Action Models",
  "eventType": "policy_or_regulation",
  "priority": "P1",
  "status": "single-source-record",
  "humanBrief": "Robot policy learning benefits from world-action models that capture environment dynamics, but pixel-level prediction entangles dynamics with nuisance factors such as lighting and texture, making learned representations vulnerable to task-irrelevant visual...",
  "machineSummary": "P1: world-state context; source payload only, no independent verification.",
  "whyItMatters": "Useful for monitoring entities, beats, source movement, and related records.",
  "scores": {
    "importanceScore": 15,
    "confidenceScore": 0.86,
    "noveltyScore": 0.7,
    "sourceDiversity": 1,
    "evidenceCount": 1
  },
  "editorialBoundary": {
    "sourceFacts": [
      "arXiv cs.RO published \"Point Tracking Improves World Action Models\" at 2026-05-22T17:08:37.000Z."
    ],
    "rnnSummary": "Robot policy learning benefits from world-action models that capture environment dynamics, but pixel-level prediction entangles dynamics with nuisance factors such as lighting and texture, making learned representations vulnerable to task-irrelevant visual...",
    "rnnInference": [
      "Event type classified as policy_or_regulation from title, tags, source type, and entities.",
      "Priority P1 is a routing readout, not a claim that the event is objectively most important."
    ],
    "uncertainOrNotVerified": [
      "RNN does not republish or inspect full copyrighted article bodies for this event record.",
      "The record is derived from source metadata, source-provided excerpts, timestamps, links, and RNN summaries.",
      "Only one source is present in this event record."
    ],
    "updatePolicy": "New source items should append evidence and updateHistory rather than overwriting source facts."
  },
  "entities": [
    {
      "name": "arXiv",
      "slug": "arxiv",
      "type": "primary-source",
      "mentions": 1
    },
    {
      "name": "robotics research",
      "slug": "robotics-research",
      "type": "surface",
      "mentions": 1
    }
  ],
  "tags": [
    "robotics-research",
    "Robotics",
    "cs.RO"
  ],
  "evidence": [
    {
      "articleId": "arxiv-query-cat-cs-ro-468314c4:veIIHZa84EBLqNighx",
      "title": "Point Tracking Improves World Action Models",
      "sourceName": "arXiv cs.RO",
      "sourceUrl": "https://arxiv.org/abs/2605.23856v1",
      "publishedAt": "2026-05-22T17:08:37.000Z",
      "confidenceScore": 0.78,
      "sourceStatus": "source-linked",
      "sourceBasis": "primary",
      "html": "/story/arxiv-query-cat-cs-ro-468314c4%3AveIIHZa84EBLqNighx",
      "json": "/api/v1/articles/arxiv-query-cat-cs-ro-468314c4%3AveIIHZa84EBLqNighx",
      "sourceFact": "arXiv cs.RO published \"Point Tracking Improves World Action Models\" at 2026-05-22T17:08:37.000Z."
    }
  ],
  "updateHistory": [
    {
      "at": "2026-05-22T17:08:37.000Z",
      "articleId": "arxiv-query-cat-cs-ro-468314c4:veIIHZa84EBLqNighx",
      "note": "Observed source metadata from arXiv cs.RO."
    }
  ],
  "canonicalSourceUrl": "https://arxiv.org/abs/2605.23856v1",
  "publishedAt": "2026-05-22T17:08:37.000Z",
  "updatedAt": "2026-05-22T17:08:37.000Z",
  "links": {
    "html": "/event-records#event-single-arxiv-query-cat-cs-ro-468314c4-veiihza84eblqnighx",
    "json": "/api/v1/events/event-single-arxiv-query-cat-cs-ro-468314c4-veiihza84eblqnighx",
    "leadStory": "/story/arxiv-query-cat-cs-ro-468314c4%3AveIIHZa84EBLqNighx",
    "leadPacket": "/api/v1/articles/arxiv-query-cat-cs-ro-468314c4%3AveIIHZa84EBLqNighx",
    "source": "https://arxiv.org/abs/2605.23856v1"
  }
}