{
  "$schema": "https://themachinepress.com/schemas/story-v1.schema.json",
  "schema_version": "1.0.0",
  "document_type": "machine_press_story",
  "story": {
    "story_id": "mp-2026-09-20-006",
    "source_story_id": "tmp-story-transgrid-systematic-generalization",
    "edition_id": "mp-2026-09-20-morning-0073",
    "edition_url": "https://themachinepress.com/edition/2026-09-20",
    "position": 6,
    "story_type": "dispatch",
    "section": "benchmarks-evals",
    "editorial_classification": "editorial",
    "headline": "The Hardest Compositional Test Fell to 15.8%",
    "slug": "the-hardest-compositional-test-fell-to-15-8",
    "dek": "TranSGrid combined deduction, induction and abduction instead of simplifying two of them away.",
    "summary": "TranSGrid combined deduction, induction and abduction instead of simplifying two of them away.",
    "body_text": "TranSGrid tests whether models can recombine known pieces when the task requires deductive, inductive and abductive reasoning together. Seven Transformers performed much worse than on an ordinary held-out set: the largest solved 79.6% of that set, 55.3% of TranSGrid and 15.8% of its hardest subset. Making actions nearly linear or making goals action-explicit returned scores toward the simpler held-out level. The result argues that familiar systematic-generalization tasks can hide the reasoning demands they claim to measure.",
    "why_it_matters": "TranSGrid combined deduction, induction and abduction instead of simplifying two of them away.",
    "limitations": [],
    "importance": 8,
    "canonical_url": "https://themachinepress.com/story/mp-2026-09-20-006/the-hardest-compositional-test-fell-to-15-8",
    "json_url": "https://themachinepress.com/story/mp-2026-09-20-006.json",
    "first_published_at": "2026-09-20T09:00:00.000-04:00",
    "modified_at": "2026-09-20T09:00:00.000-04:00",
    "content_status": "new",
    "is_carryover": false,
    "carryover_reason": null,
    "key_claims": [
      {
        "claim_id": "claim-mp-2026-09-20-006-001",
        "text": "TranSGrid combined deduction, induction and abduction instead of simplifying two of them away.",
        "source_ids": [
          "source-2026-09-20-006"
        ],
        "qualification": null
      }
    ],
    "source_ids": [
      "source-2026-09-20-006"
    ],
    "tags": [
      "systematic generalization",
      "reasoning",
      "benchmarks"
    ],
    "image_url": null,
    "corrections": []
  },
  "sources": [
    {
      "source_id": "source-2026-09-20-006",
      "title": "arXiv preprint 2609.19212",
      "publisher": "arXiv",
      "url": "https://arxiv.org/abs/2609.19212",
      "canonical_url": "https://arxiv.org/abs/2609.19212",
      "source_type": "primary_research",
      "is_primary_source": true,
      "published_at": "2026-09-16T07:53:03.000-04:00",
      "accessed_at": "2026-09-20T08:23:59.913-04:00",
      "supports_claim_ids": [
        "claim-mp-2026-09-20-006-001"
      ]
    }
  ],
  "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 Hardest Compositional Test Fell to 15.8%",
    "publisher": "The Machine Press",
    "published_at": "2026-09-20T09:00:00.000-04:00",
    "canonical_url": "https://themachinepress.com/story/mp-2026-09-20-006/the-hardest-compositional-test-fell-to-15-8"
  }
}