{
  "$schema": "https://themachinepress.com/schemas/story-v1.schema.json",
  "schema_version": "1.0.0",
  "document_type": "machine_press_story",
  "story": {
    "story_id": "mp-2026-09-21-013",
    "source_story_id": "tmp-story-vla-feedback-denoising",
    "edition_id": "mp-2026-09-21-morning-0074",
    "edition_url": "https://themachinepress.com/edition/2026-09-21",
    "position": 13,
    "story_type": "dispatch",
    "section": "robotics",
    "editorial_classification": "editorial",
    "headline": "The Robot Corrected an Action Before It Finished",
    "slug": "the-robot-corrected-an-action-before-it-finished",
    "dek": "A light final denoising step brought fresh visual feedback into an existing action chunk.",
    "summary": "A light final denoising step brought fresh visual feedback into an existing action chunk.",
    "body_text": "VLA-Feedback pairs a slower diffusion planner with a fast visual correction at the final denoising step of each action. Rather than rerun the full model whenever the scene changes, it updates the pending action using the newest observation. The authors report the same performance as GR00T on static LIBERO tasks, while dynamic simulation success rose from 27.5% to 85.0% and average real-robot success from 51% to 73%. These are results in the reported tasks and comparison setup, not general deployment performance.",
    "why_it_matters": "A light final denoising step brought fresh visual feedback into an existing action chunk.",
    "limitations": [
      "These are results in the reported tasks and comparison setup, not general deployment performance."
    ],
    "importance": 9,
    "canonical_url": "https://themachinepress.com/story/mp-2026-09-21-013/the-robot-corrected-an-action-before-it-finished",
    "json_url": "https://themachinepress.com/story/mp-2026-09-21-013.json",
    "first_published_at": "2026-09-21T09:00:00.000-04:00",
    "modified_at": "2026-09-21T09:00:00.000-04:00",
    "content_status": "new",
    "is_carryover": false,
    "carryover_reason": null,
    "key_claims": [
      {
        "claim_id": "claim-mp-2026-09-21-013-001",
        "text": "A light final denoising step brought fresh visual feedback into an existing action chunk.",
        "source_ids": [
          "source-2026-09-21-013"
        ],
        "qualification": "These are results in the reported tasks and comparison setup, not general deployment performance."
      }
    ],
    "source_ids": [
      "source-2026-09-21-013"
    ],
    "tags": [
      "VLA",
      "visual feedback",
      "robot manipulation"
    ],
    "image_url": "https://themachinepress.com/issues/2026-09-21/vla-code-file.webp",
    "corrections": []
  },
  "sources": [
    {
      "source_id": "source-2026-09-21-013",
      "title": "arXiv preprint 2609.21022",
      "publisher": "arXiv",
      "url": "https://arxiv.org/abs/2609.21022",
      "canonical_url": "https://arxiv.org/abs/2609.21022",
      "source_type": "primary_research",
      "is_primary_source": true,
      "published_at": "2026-09-17T15:20:15.000-04:00",
      "accessed_at": "2026-09-21T08:26:05.112-04:00",
      "supports_claim_ids": [
        "claim-mp-2026-09-21-013-001"
      ]
    }
  ],
  "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 Corrected an Action Before It Finished",
    "publisher": "The Machine Press",
    "published_at": "2026-09-21T09:00:00.000-04:00",
    "canonical_url": "https://themachinepress.com/story/mp-2026-09-21-013/the-robot-corrected-an-action-before-it-finished"
  }
}