{
  "$schema": "https://themachinepress.com/schemas/story-v1.schema.json",
  "schema_version": "1.0.0",
  "document_type": "machine_press_story",
  "story": {
    "story_id": "mp-2026-09-09-006",
    "source_story_id": "tmp-story-layerroute-vla-representation-routing",
    "edition_id": "mp-2026-09-09-morning-0062",
    "edition_url": "https://themachinepress.com/edition/2026-09-09",
    "position": 6,
    "story_type": "dispatch",
    "section": "robotics",
    "editorial_classification": "editorial",
    "headline": "The Robot Chose Which Layer to Remember Mid-Action",
    "slug": "the-robot-chose-which-layer-to-remember-mid-action",
    "dek": "LayerRoute dynamically mixes visual-language layers and rereads earlier action states, adding as little as 0.31 percent in one tested policy.",
    "summary": "LayerRoute dynamically mixes visual-language layers and rereads earlier action states, adding as little as 0.31 percent in one tested policy.",
    "body_text": "Vision-language-action policies commonly expose fixed visual-language layers to each action layer and leave intermediate action states implicit. LayerRoute adds two interfaces: a router that mixes cached visual-language representations according to the current action state, and a reread path for earlier action representations. Across simulation and real-robot benchmarks, the authors report consistent gains for two base policies, including up to 7.2 points on LIBERO Long with 0.31 percent and 3.87 percent additional parameters in the respective systems. The results support adaptive representation access, but do not establish a universal routing recipe for every robot or task.",
    "why_it_matters": "LayerRoute dynamically mixes visual-language layers and rereads earlier action states, adding as little as 0.31 percent in one tested policy.",
    "limitations": [
      "LayerRoute adds two interfaces: a router that mixes cached visual-language representations according to the current action state, and a reread path for earlier action representations.",
      "The results support adaptive representation access, but do not establish a universal routing recipe for every robot or task."
    ],
    "importance": 8,
    "canonical_url": "https://themachinepress.com/story/mp-2026-09-09-006/the-robot-chose-which-layer-to-remember-mid-action",
    "json_url": "https://themachinepress.com/story/mp-2026-09-09-006.json",
    "first_published_at": "2026-09-09T09:00:00.000-04:00",
    "modified_at": "2026-09-09T09:00:00.000-04:00",
    "content_status": "new",
    "is_carryover": false,
    "carryover_reason": null,
    "key_claims": [
      {
        "claim_id": "claim-mp-2026-09-09-006-001",
        "text": "LayerRoute dynamically mixes visual-language layers and rereads earlier action states, adding as little as 0.31 percent in one tested policy.",
        "source_ids": [
          "source-2026-09-09-006"
        ],
        "qualification": "LayerRoute adds two interfaces: a router that mixes cached visual-language representations according to the current action state, and a reread path for earlier action representations."
      }
    ],
    "source_ids": [
      "source-2026-09-09-006"
    ],
    "tags": [
      "vision-language-action",
      "robot manipulation",
      "representation routing"
    ],
    "image_url": null,
    "corrections": []
  },
  "sources": [
    {
      "source_id": "source-2026-09-09-006",
      "title": "arXiv preprint 2609.06079",
      "publisher": "arXiv",
      "url": "https://arxiv.org/abs/2609.06079",
      "canonical_url": "https://arxiv.org/abs/2609.06079",
      "source_type": "primary_research",
      "is_primary_source": true,
      "published_at": "2026-09-05T09:13:07.000-04:00",
      "accessed_at": "2026-09-09T08:30:00.000-04:00",
      "supports_claim_ids": [
        "claim-mp-2026-09-09-006-001"
      ]
    }
  ],
  "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 Chose Which Layer to Remember Mid-Action",
    "publisher": "The Machine Press",
    "published_at": "2026-09-09T09:00:00.000-04:00",
    "canonical_url": "https://themachinepress.com/story/mp-2026-09-09-006/the-robot-chose-which-layer-to-remember-mid-action"
  }
}