{
  "$schema": "https://themachinepress.com/schemas/story-v1.schema.json",
  "schema_version": "1.0.0",
  "document_type": "machine_press_story",
  "story": {
    "story_id": "mp-2026-11-04-010",
    "source_story_id": "tmp-story-medphysbench-error-detection",
    "edition_id": "mp-2026-11-04-morning-0117",
    "edition_url": "https://themachinepress.com/edition/2026-11-04",
    "position": 10,
    "story_type": "dispatch",
    "section": "benchmarks-evals",
    "editorial_classification": "editorial",
    "headline": "The Model Can Calculate Correctly and Still Miss the Wrong Worksheet",
    "slug": "the-model-can-calculate-correctly-and-still-miss-the-wrong-worksheet",
    "dek": "MedPhysBench separates medical-physics calculation from error detection, exposing uneven performance on clinical tasks and planted mistakes.",
    "summary": "MedPhysBench separates medical-physics calculation from error detection, exposing uneven performance on clinical tasks and planted mistakes.",
    "body_text": "MedPhysBench generates calculation and checking items across radiation therapy, imaging, nuclear medicine and radiation protection. On 1,200 held-out items, 12 models ranged from 33.3 to 100 percent calculation accuracy; the authors also found problems in two shielding items and disclose them. Five models caught at least 96.7 percent of mostly large planted errors with no more than 4 percent false alarms. In a targeted inverse-square test, some weaker models performed the correction in their own work yet accepted most worksheets that omitted it. The benchmark is a research evaluation with synthetic and public cases, not authorization for unsupervised clinical use.",
    "why_it_matters": "MedPhysBench separates medical-physics calculation from error detection, exposing uneven performance on clinical tasks and planted mistakes.",
    "limitations": [
      "The benchmark is a research evaluation with synthetic and public cases, not authorization for unsupervised clinical use."
    ],
    "importance": 9,
    "canonical_url": "https://themachinepress.com/story/mp-2026-11-04-010/the-model-can-calculate-correctly-and-still-miss-the-wrong-worksheet",
    "json_url": "https://themachinepress.com/story/mp-2026-11-04-010.json",
    "first_published_at": "2026-11-04T09:00:00.000-05:00",
    "modified_at": "2026-11-04T09:00:00.000-05:00",
    "content_status": "new",
    "is_carryover": false,
    "carryover_reason": null,
    "key_claims": [
      {
        "claim_id": "claim-mp-2026-11-04-010-001",
        "text": "MedPhysBench separates medical-physics calculation from error detection, exposing uneven performance on clinical tasks and planted mistakes.",
        "source_ids": [
          "source-2026-11-04-010"
        ],
        "qualification": "The benchmark is a research evaluation with synthetic and public cases, not authorization for unsupervised clinical use."
      }
    ],
    "source_ids": [
      "source-2026-11-04-010"
    ],
    "tags": [
      "medical physics",
      "benchmarks",
      "error detection"
    ],
    "image_url": null,
    "corrections": []
  },
  "sources": [
    {
      "source_id": "source-2026-11-04-010",
      "title": "arXiv preprint 2611.05340, original v1",
      "publisher": "arXiv",
      "url": "https://arxiv.org/abs/2611.05340v1",
      "canonical_url": "https://arxiv.org/abs/2611.05340v1",
      "source_type": "primary_research",
      "is_primary_source": true,
      "published_at": "2026-11-03T13:40:35.000-05:00",
      "accessed_at": "2026-11-04T08:28:27.560-05:00",
      "supports_claim_ids": [
        "claim-mp-2026-11-04-010-001"
      ]
    }
  ],
  "corrections": [],
  "publisher": {
    "name": "The Machine Press",
    "url": "https://themachinepress.com",
    "description": "A daily newspaper for the age of artificial intelligence."
  },
  "cite_this_report": {
    "title": "The Model Can Calculate Correctly and Still Miss the Wrong Worksheet",
    "publisher": "The Machine Press",
    "published_at": "2026-11-04T09:00:00.000-05:00",
    "canonical_url": "https://themachinepress.com/story/mp-2026-11-04-010/the-model-can-calculate-correctly-and-still-miss-the-wrong-worksheet"
  }
}