{
  "$schema": "https://themachinepress.com/schemas/story-v1.schema.json",
  "schema_version": "1.0.0",
  "document_type": "machine_press_story",
  "story": {
    "story_id": "mp-2026-09-30-026",
    "source_story_id": "tmp-story-egogears-cross-encounter-video",
    "edition_id": "mp-2026-09-30-morning-0083",
    "edition_url": "https://themachinepress.com/edition/2026-09-30",
    "position": 15,
    "story_type": "dispatch",
    "section": "benchmarks-evals",
    "editorial_classification": "editorial",
    "headline": "Every Comparable Model Lost Ground Across Repeated Encounters",
    "slug": "every-comparable-model-lost-ground-across-repeated-encounters",
    "dek": "EgoGears separated local perception from the harder job of binding evidence across independent route videos.",
    "summary": "EgoGears separated local perception from the harder job of binding evidence across independent route videos.",
    "body_text": "EgoGears contains 567 single-video and 1,487 multi-video questions from 126 egocentric recordings of 39 outdoor routes; 531 questions require alignment across recordings. Among 20 model configurations tested comparably on both splits, every one performed worse on multi-video questions, with a mean decline of 22.5 percentage points. Controls indicate that additional video volume or recording boundaries alone do not explain the gap. The authors identify observation-evidence binding and ordered route-state tracking as the central tested bottlenecks.",
    "why_it_matters": "EgoGears separated local perception from the harder job of binding evidence across independent route videos.",
    "limitations": [
      "Controls indicate that additional video volume or recording boundaries alone do not explain the gap."
    ],
    "importance": 8,
    "canonical_url": "https://themachinepress.com/story/mp-2026-09-30-026/every-comparable-model-lost-ground-across-repeated-encounters",
    "json_url": "https://themachinepress.com/story/mp-2026-09-30-026.json",
    "first_published_at": "2026-09-30T09:00:00.000-04:00",
    "modified_at": "2026-09-30T09:02:00.000-04:00",
    "content_status": "new",
    "is_carryover": false,
    "carryover_reason": null,
    "key_claims": [
      {
        "claim_id": "claim-mp-2026-09-30-026-001",
        "text": "EgoGears separated local perception from the harder job of binding evidence across independent route videos.",
        "source_ids": [
          "source-2026-09-30-015"
        ],
        "qualification": "Controls indicate that additional video volume or recording boundaries alone do not explain the gap."
      }
    ],
    "source_ids": [
      "source-2026-09-30-015"
    ],
    "tags": [
      "video benchmarks",
      "egocentric vision",
      "cross-encounter reasoning"
    ],
    "image_url": null,
    "corrections": []
  },
  "sources": [
    {
      "source_id": "source-2026-09-30-015",
      "title": "arXiv preprint 2609.37938",
      "publisher": "arXiv",
      "url": "https://arxiv.org/abs/2609.37938",
      "canonical_url": "https://arxiv.org/abs/2609.37938",
      "source_type": "primary_research",
      "is_primary_source": true,
      "published_at": "2026-09-29T12:20:01.000-04:00",
      "accessed_at": "2026-09-30T08:24:00.000-04:00",
      "supports_claim_ids": [
        "claim-mp-2026-09-30-026-001"
      ]
    }
  ],
  "corrections": [],
  "publisher": {
    "name": "The Machine Press",
    "url": "https://themachinepress.com",
    "description": "A daily newspaper for the age of artificial intelligence."
  },
  "cite_this_report": {
    "title": "Every Comparable Model Lost Ground Across Repeated Encounters",
    "publisher": "The Machine Press",
    "published_at": "2026-09-30T09:00:00.000-04:00",
    "canonical_url": "https://themachinepress.com/story/mp-2026-09-30-026/every-comparable-model-lost-ground-across-repeated-encounters"
  }
}