{
  "$schema": "https://themachinepress.com/schemas/story-v1.schema.json",
  "schema_version": "1.0.0",
  "document_type": "machine_press_story",
  "story": {
    "story_id": "mp-2026-09-23-012",
    "source_story_id": "tmp-story-robofollow-instruction-mirage",
    "edition_id": "mp-2026-09-23-morning-0076",
    "edition_url": "https://themachinepress.com/edition/2026-09-23",
    "position": 12,
    "story_type": "dispatch",
    "section": "benchmarks-evals",
    "editorial_classification": "editorial",
    "headline": "Robot Success Hid Weak Instruction Following",
    "slug": "robot-success-hid-weak-instruction-following",
    "dek": "Scenes with only one plausible task let embodied policies ignore language.",
    "summary": "Scenes with only one plausible task let embodied policies ignore language.",
    "body_text": "RoboFollow raises scene entropy by placing multiple valid task branches in one scene, so a policy must distinguish instructions rather than infer the sole available action. Nine evaluated VLA and world-action models that performed well in the easiest setting did not reliably transfer across progressively changed layouts and semantics after fine-tuning. Stronger vision-language backbones and several training adjustments did not close the gap. This is a diagnostic benchmark finding, not a claim that every deployed robot ignores language.",
    "why_it_matters": "Scenes with only one plausible task let embodied policies ignore language.",
    "limitations": [
      "Nine evaluated VLA and world-action models that performed well in the easiest setting did not reliably transfer across progressively changed layouts and semantics after fine-tuning.",
      "Stronger vision-language backbones and several training adjustments did not close the gap.",
      "This is a diagnostic benchmark finding, not a claim that every deployed robot ignores language."
    ],
    "importance": 8,
    "canonical_url": "https://themachinepress.com/story/mp-2026-09-23-012/robot-success-hid-weak-instruction-following",
    "json_url": "https://themachinepress.com/story/mp-2026-09-23-012.json",
    "first_published_at": "2026-09-23T09:00:00.000-04:00",
    "modified_at": "2026-09-23T09:00:00.000-04:00",
    "content_status": "new",
    "is_carryover": false,
    "carryover_reason": null,
    "key_claims": [
      {
        "claim_id": "claim-mp-2026-09-23-012-001",
        "text": "Scenes with only one plausible task let embodied policies ignore language.",
        "source_ids": [
          "source-2026-09-23-012"
        ],
        "qualification": "Nine evaluated VLA and world-action models that performed well in the easiest setting did not reliably transfer across progressively changed layouts and semantics after fine-tuning."
      }
    ],
    "source_ids": [
      "source-2026-09-23-012"
    ],
    "tags": [
      "embodied agents",
      "instruction following",
      "benchmarks"
    ],
    "image_url": null,
    "corrections": []
  },
  "sources": [
    {
      "source_id": "source-2026-09-23-012",
      "title": "arXiv preprint 2609.25636",
      "publisher": "arXiv",
      "url": "https://arxiv.org/abs/2609.25636",
      "canonical_url": "https://arxiv.org/abs/2609.25636",
      "source_type": "primary_research",
      "is_primary_source": true,
      "published_at": "2026-09-21T23:48:02.000-04:00",
      "accessed_at": "2026-09-23T08:22:43.661-04:00",
      "supports_claim_ids": [
        "claim-mp-2026-09-23-012-001"
      ]
    }
  ],
  "corrections": [],
  "publisher": {
    "name": "The Machine Press",
    "url": "https://themachinepress.com",
    "description": "A daily newspaper for the age of artificial intelligence."
  },
  "cite_this_report": {
    "title": "Robot Success Hid Weak Instruction Following",
    "publisher": "The Machine Press",
    "published_at": "2026-09-23T09:00:00.000-04:00",
    "canonical_url": "https://themachinepress.com/story/mp-2026-09-23-012/robot-success-hid-weak-instruction-following"
  }
}