{
  "$schema": "https://themachinepress.com/schemas/story-v1.schema.json",
  "schema_version": "1.0.0",
  "document_type": "machine_press_story",
  "story": {
    "story_id": "mp-2026-10-23-014",
    "source_story_id": "tmp-story-progress-probes-vla-monitoring",
    "edition_id": "mp-2026-10-23-morning-0105",
    "edition_url": "https://themachinepress.com/edition/2026-10-23",
    "position": 14,
    "story_type": "dispatch",
    "section": "robotics",
    "editorial_classification": "editorial",
    "headline": "The Robot Policy Watches Its Own Progress",
    "slug": "the-robot-policy-watches-its-own-progress",
    "dek": "Linear probes trained on internal activations flag executions that depart from the progress pattern seen in successful demonstrations.",
    "summary": "Linear probes trained on internal activations flag executions that depart from the progress pattern seen in successful demonstrations.",
    "body_text": "Runtime monitors for vision-language-action policies often add another foundation model to watch the first. Progress Probes instead trains lightweight linear probes on the policy's own activations from demonstrations. The monitor estimates how far a task should have advanced and flags trajectories that diverge from that expected progress. In simulated and real manipulation tests with the open-source pi-0.5 policy, the probes matched or exceeded vision-language-model monitors and approaches that require failure examples. The evidence covers the tested policy and tasks; internal progress signals do not guarantee detection of every failure mode.",
    "why_it_matters": "Linear probes trained on internal activations flag executions that depart from the progress pattern seen in successful demonstrations.",
    "limitations": [
      "The evidence covers the tested policy and tasks; internal progress signals do not guarantee detection of every failure mode."
    ],
    "importance": 8,
    "canonical_url": "https://themachinepress.com/story/mp-2026-10-23-014/the-robot-policy-watches-its-own-progress",
    "json_url": "https://themachinepress.com/story/mp-2026-10-23-014.json",
    "first_published_at": "2026-10-23T09:00:00.000-04:00",
    "modified_at": "2026-10-23T09:00:00.000-04:00",
    "content_status": "new",
    "is_carryover": false,
    "carryover_reason": null,
    "key_claims": [
      {
        "claim_id": "claim-mp-2026-10-23-014-001",
        "text": "Linear probes trained on internal activations flag executions that depart from the progress pattern seen in successful demonstrations.",
        "source_ids": [
          "source-2026-10-23-014"
        ],
        "qualification": "The evidence covers the tested policy and tasks; internal progress signals do not guarantee detection of every failure mode."
      }
    ],
    "source_ids": [
      "source-2026-10-23-014"
    ],
    "tags": [
      "robotics",
      "runtime monitoring",
      "VLA"
    ],
    "image_url": null,
    "corrections": []
  },
  "sources": [
    {
      "source_id": "source-2026-10-23-014",
      "title": "arXiv preprint 2610.31613, original v1",
      "publisher": "arXiv",
      "url": "https://arxiv.org/abs/2610.31613v1",
      "canonical_url": "https://arxiv.org/abs/2610.31613v1",
      "source_type": "primary_research",
      "is_primary_source": true,
      "published_at": "2026-10-22T13:40:19.000-04:00",
      "accessed_at": "2026-10-23T08:21:38.024-04:00",
      "supports_claim_ids": [
        "claim-mp-2026-10-23-014-001"
      ]
    }
  ],
  "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 Policy Watches Its Own Progress",
    "publisher": "The Machine Press",
    "published_at": "2026-10-23T09:00:00.000-04:00",
    "canonical_url": "https://themachinepress.com/story/mp-2026-10-23-014/the-robot-policy-watches-its-own-progress"
  }
}