{
  "$schema": "https://themachinepress.com/schemas/story-v1.schema.json",
  "schema_version": "1.0.0",
  "document_type": "machine_press_story",
  "story": {
    "story_id": "mp-2026-10-18-013",
    "source_story_id": "tmp-story-mem4act-latent-memory",
    "edition_id": "mp-2026-10-18-morning-0100",
    "edition_url": "https://themachinepress.com/edition/2026-10-18",
    "position": 13,
    "story_type": "dispatch",
    "section": "robotics",
    "editorial_classification": "editorial",
    "headline": "The Robot Compresses Its Past Into a Fixed-Size Memory",
    "slug": "the-robot-compresses-its-past-into-a-fixed-size-memory",
    "dek": "Mem4Act supervises semantic and perceptual recall, then trains the action policy to depend on that bounded state.",
    "summary": "Mem4Act supervises semantic and perceptual recall, then trains the action policy to depend on that bounded state.",
    "body_text": "Long-horizon robot policies can keep raw frames, compressed tokens or language summaries, but each choice either lengthens context or risks losing the detail needed later. Mem4Act aggregates history at multiple timescales, supervises memory through question answering and key-frame feature reconstruction, and uses memory-first training so the action policy cannot simply ignore the compressed state. The authors report success rates of 49.88 percent on RoboMME, 61.0 percent on RMBench and 81.5 percent across four real-world manipulation tasks. The evaluation is bounded to those tasks, yet it demonstrates a practical memory interface whose size does not grow with the episode.",
    "why_it_matters": "Mem4Act supervises semantic and perceptual recall, then trains the action policy to depend on that bounded state.",
    "limitations": [
      "Mem4Act aggregates history at multiple timescales, supervises memory through question answering and key-frame feature reconstruction, and uses memory-first training so the action policy cannot simply ignore the compressed state.",
      "The evaluation is bounded to those tasks, yet it demonstrates a practical memory interface whose size does not grow with the episode."
    ],
    "importance": 8,
    "canonical_url": "https://themachinepress.com/story/mp-2026-10-18-013/the-robot-compresses-its-past-into-a-fixed-size-memory",
    "json_url": "https://themachinepress.com/story/mp-2026-10-18-013.json",
    "first_published_at": "2026-10-18T09:00:00.000-04:00",
    "modified_at": "2026-10-18T09:00:00.000-04:00",
    "content_status": "new",
    "is_carryover": false,
    "carryover_reason": null,
    "key_claims": [
      {
        "claim_id": "claim-mp-2026-10-18-013-001",
        "text": "Mem4Act supervises semantic and perceptual recall, then trains the action policy to depend on that bounded state.",
        "source_ids": [
          "source-2026-10-18-013"
        ],
        "qualification": "Mem4Act aggregates history at multiple timescales, supervises memory through question answering and key-frame feature reconstruction, and uses memory-first training so the action policy cannot simply ignore the compressed state."
      }
    ],
    "source_ids": [
      "source-2026-10-18-013"
    ],
    "tags": [
      "robotics",
      "memory",
      "long-horizon control"
    ],
    "image_url": "https://themachinepress.com/issues/2026-10-18/mem4act-robot.webp",
    "corrections": []
  },
  "sources": [
    {
      "source_id": "source-2026-10-18-013",
      "title": "arXiv preprint 2610.22323, original v1",
      "publisher": "arXiv",
      "url": "https://arxiv.org/abs/2610.22323v1",
      "canonical_url": "https://arxiv.org/abs/2610.22323v1",
      "source_type": "primary_research",
      "is_primary_source": true,
      "published_at": "2026-10-15T13:51:58.000-04:00",
      "accessed_at": "2026-10-18T08:23:52.920-04:00",
      "supports_claim_ids": [
        "claim-mp-2026-10-18-013-001"
      ]
    }
  ],
  "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 Robot Compresses Its Past Into a Fixed-Size Memory",
    "publisher": "The Machine Press",
    "published_at": "2026-10-18T09:00:00.000-04:00",
    "canonical_url": "https://themachinepress.com/story/mp-2026-10-18-013/the-robot-compresses-its-past-into-a-fixed-size-memory"
  }
}