{
  "$schema": "https://themachinepress.com/schemas/story-v1.schema.json",
  "schema_version": "1.0.0",
  "document_type": "machine_press_story",
  "story": {
    "story_id": "mp-2026-08-24-008",
    "source_story_id": "tmp-story-flavourbench-executable-ground-truth",
    "edition_id": "mp-2026-08-24-morning-0046",
    "edition_url": "https://themachinepress.com/edition/2026-08-24",
    "position": 8,
    "story_type": "dispatch",
    "section": "benchmarks-evals",
    "editorial_classification": "editorial",
    "headline": "The Culinary Judge Scored Every Possible Plate Before the Models Arrived",
    "slug": "the-culinary-judge-scored-every-possible-plate-before-the-models-arrived",
    "dek": "FlavourBench freezes executable scores for all 56 three-ingredient portfolios in each task.",
    "summary": "FlavourBench freezes executable scores for all 56 three-ingredient portfolios in each task.",
    "body_text": "The benchmark evaluated 27 frontier endpoints on 534 identical tasks, totaling 14,418 scored model-task cells without differential missingness. Its two independently compiled panels correlated at 0.89, and 101 of 351 model pairs were statistically resolved; the authors release prompts, raw responses, score maps and an offline verifier.",
    "why_it_matters": "FlavourBench freezes executable scores for all 56 three-ingredient portfolios in each task.",
    "limitations": [],
    "importance": 8,
    "canonical_url": "https://themachinepress.com/story/mp-2026-08-24-008/the-culinary-judge-scored-every-possible-plate-before-the-models-arrived",
    "json_url": "https://themachinepress.com/story/mp-2026-08-24-008.json",
    "first_published_at": "2026-08-24T09:00:00.000-04:00",
    "modified_at": "2026-08-24T09:00:00.000-04:00",
    "content_status": "new",
    "is_carryover": false,
    "carryover_reason": null,
    "key_claims": [
      {
        "claim_id": "claim-mp-2026-08-24-008-001",
        "text": "FlavourBench freezes executable scores for all 56 three-ingredient portfolios in each task.",
        "source_ids": [
          "source-2026-08-24-008"
        ],
        "qualification": null
      }
    ],
    "source_ids": [
      "source-2026-08-24-008"
    ],
    "tags": [
      "benchmarking",
      "executable evaluation",
      "language models"
    ],
    "image_url": "https://themachinepress.com/issues/2026-08-24/flavourbench-code-file-image.webp",
    "corrections": []
  },
  "sources": [
    {
      "source_id": "source-2026-08-24-008",
      "title": "arXiv preprint 2608.20574",
      "publisher": "arXiv",
      "url": "https://arxiv.org/abs/2608.20574",
      "canonical_url": "https://arxiv.org/abs/2608.20574",
      "source_type": "primary_research",
      "is_primary_source": true,
      "published_at": "2026-08-20T17:14:52.000-04:00",
      "accessed_at": "2026-08-24T08:23:00.000-04:00",
      "supports_claim_ids": [
        "claim-mp-2026-08-24-008-001"
      ]
    }
  ],
  "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 Culinary Judge Scored Every Possible Plate Before the Models Arrived",
    "publisher": "The Machine Press",
    "published_at": "2026-08-24T09:00:00.000-04:00",
    "canonical_url": "https://themachinepress.com/story/mp-2026-08-24-008/the-culinary-judge-scored-every-possible-plate-before-the-models-arrived"
  }
}