{
  "$schema": "https://themachinepress.com/schemas/story-v1.schema.json",
  "schema_version": "1.0.0",
  "document_type": "machine_press_story",
  "story": {
    "story_id": "mp-2026-09-17-010",
    "source_story_id": "tmp-story-forcedelta-contact-corrections",
    "edition_id": "mp-2026-09-17-morning-0070",
    "edition_url": "https://themachinepress.com/edition/2026-09-17",
    "position": 10,
    "story_type": "dispatch",
    "section": "robotics",
    "editorial_classification": "editorial",
    "headline": "Force Corrections Lifted Contact-Rich Success to 82.2%",
    "slug": "force-corrections-lifted-contact-rich-success-to-82-2",
    "dek": "A lightweight correction policy beat the 54.4% ForceVLA baseline and cut peak force by about 26%.",
    "summary": "A lightweight correction policy beat the 54.4% ForceVLA baseline and cut peak force by about 26%.",
    "body_text": "ForceDelta-VLA separates a task-level reference action from contact-dependent corrections. A frozen teacher supplies paired force-aware and force-agnostic predictions, allowing the system to distill an explicit correction target without labeled decomposition. Across nine single-arm and bimanual tasks, the full policy reported 82.2% mean success, against 54.4% for ForceVLA and 70.6% for the first-stage temporal teacher. On successful trials, mean peak contact force fell by roughly 26% on both platforms. The correction policy can react between slower reference-action updates using recent force history.",
    "why_it_matters": "A lightweight correction policy beat the 54.4% ForceVLA baseline and cut peak force by about 26%.",
    "limitations": [],
    "importance": 8,
    "canonical_url": "https://themachinepress.com/story/mp-2026-09-17-010/force-corrections-lifted-contact-rich-success-to-82-2",
    "json_url": "https://themachinepress.com/story/mp-2026-09-17-010.json",
    "first_published_at": "2026-09-17T09:00:00.000-04:00",
    "modified_at": "2026-09-17T09:00:00.000-04:00",
    "content_status": "new",
    "is_carryover": false,
    "carryover_reason": null,
    "key_claims": [
      {
        "claim_id": "claim-mp-2026-09-17-010-001",
        "text": "A lightweight correction policy beat the 54.4% ForceVLA baseline and cut peak force by about 26%.",
        "source_ids": [
          "source-2026-09-17-010"
        ],
        "qualification": null
      }
    ],
    "source_ids": [
      "source-2026-09-17-010"
    ],
    "tags": [
      "contact-rich manipulation",
      "force feedback",
      "robot learning"
    ],
    "image_url": null,
    "corrections": []
  },
  "sources": [
    {
      "source_id": "source-2026-09-17-010",
      "title": "arXiv preprint 2609.18242",
      "publisher": "arXiv",
      "url": "https://arxiv.org/abs/2609.18242",
      "canonical_url": "https://arxiv.org/abs/2609.18242",
      "source_type": "primary_research",
      "is_primary_source": true,
      "published_at": "2026-09-16T03:20:10.000-04:00",
      "accessed_at": "2026-09-17T08:28:34.205-04:00",
      "supports_claim_ids": [
        "claim-mp-2026-09-17-010-001"
      ]
    }
  ],
  "corrections": [],
  "publisher": {
    "name": "The Machine Press",
    "url": "https://themachinepress.com",
    "description": "A daily newspaper for the age of artificial intelligence."
  },
  "cite_this_report": {
    "title": "Force Corrections Lifted Contact-Rich Success to 82.2%",
    "publisher": "The Machine Press",
    "published_at": "2026-09-17T09:00:00.000-04:00",
    "canonical_url": "https://themachinepress.com/story/mp-2026-09-17-010/force-corrections-lifted-contact-rich-success-to-82-2"
  }
}