{
  "$schema": "https://themachinepress.com/schemas/story-v1.schema.json",
  "schema_version": "1.0.0",
  "document_type": "machine_press_story",
  "story": {
    "story_id": "mp-2026-09-18-008",
    "source_story_id": "tmp-story-hil-umi-robot-free-posttraining",
    "edition_id": "mp-2026-09-18-morning-0071",
    "edition_url": "https://themachinepress.com/edition/2026-09-18",
    "position": 8,
    "story_type": "dispatch",
    "section": "robotics",
    "editorial_classification": "editorial",
    "headline": "The Human Corrected the Policy Without Running the Robot",
    "slug": "the-human-corrected-the-policy-without-running-the-robot",
    "dek": "HIL-UMI queried the current policy during handheld demonstrations and collected data where action disagreement was high.",
    "summary": "HIL-UMI queried the current policy during handheld demonstrations and collected data where action disagreement was high.",
    "body_text": "Interactive robot post-training usually means repeatedly running the policy on hardware and waiting for human intervention. HIL-UMI instead queries the current policy on the observation stream from a handheld Universal Manipulation Interface without executing the predicted motion. An energy score targets collection where human and policy trajectories disagree, while low advantage predictions identify segments that refine a progress estimator. Across four physical tasks, the policy improved over supervised fine-tuning and beat HG-DAgger on table cleanup with lower per-frame collection time.",
    "why_it_matters": "HIL-UMI queried the current policy during handheld demonstrations and collected data where action disagreement was high.",
    "limitations": [],
    "importance": 8,
    "canonical_url": "https://themachinepress.com/story/mp-2026-09-18-008/the-human-corrected-the-policy-without-running-the-robot",
    "json_url": "https://themachinepress.com/story/mp-2026-09-18-008.json",
    "first_published_at": "2026-09-18T09:00:00.000-04:00",
    "modified_at": "2026-09-18T09:00:00.000-04:00",
    "content_status": "new",
    "is_carryover": false,
    "carryover_reason": null,
    "key_claims": [
      {
        "claim_id": "claim-mp-2026-09-18-008-001",
        "text": "HIL-UMI queried the current policy during handheld demonstrations and collected data where action disagreement was high.",
        "source_ids": [
          "source-2026-09-18-008"
        ],
        "qualification": null
      }
    ],
    "source_ids": [
      "source-2026-09-18-008"
    ],
    "tags": [
      "human in the loop",
      "VLA post-training",
      "robot data"
    ],
    "image_url": "https://themachinepress.com/issues/2026-09-18/hil-umi-robot-file.webp",
    "corrections": []
  },
  "sources": [
    {
      "source_id": "source-2026-09-18-008",
      "title": "arXiv preprint 2609.20659",
      "publisher": "arXiv",
      "url": "https://arxiv.org/abs/2609.20659",
      "canonical_url": "https://arxiv.org/abs/2609.20659",
      "source_type": "primary_research",
      "is_primary_source": true,
      "published_at": "2026-09-17T12:38:37.000-04:00",
      "accessed_at": "2026-09-18T08:26:40.599-04:00",
      "supports_claim_ids": [
        "claim-mp-2026-09-18-008-001"
      ]
    }
  ],
  "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 Human Corrected the Policy Without Running the Robot",
    "publisher": "The Machine Press",
    "published_at": "2026-09-18T09:00:00.000-04:00",
    "canonical_url": "https://themachinepress.com/story/mp-2026-09-18-008/the-human-corrected-the-policy-without-running-the-robot"
  }
}