{
  "$schema": "https://themachinepress.com/schemas/story-v1.schema.json",
  "schema_version": "1.0.0",
  "document_type": "machine_press_story",
  "story": {
    "story_id": "mp-2026-08-30-015",
    "source_story_id": "tmp-sidebar-reasoning-tax",
    "edition_id": "mp-2026-08-30-morning-0052",
    "edition_url": "https://themachinepress.com/edition/2026-08-30",
    "position": 17,
    "story_type": "ticker",
    "section": "market-industry",
    "editorial_classification": "editorial",
    "headline": "Thinking Tokens Had Diminishing Returns",
    "slug": "thinking-tokens-had-diminishing-returns",
    "dek": "A 151-run analysis finds task structure predicts marginal reasoning efficiency better than nominal difficulty.",
    "summary": "A 151-run analysis finds task structure predicts marginal reasoning efficiency better than nominal difficulty.",
    "body_text": "Sequential inference tasks showed stronger token-normalized gains than knowledge recall, while higher reasoning effort sometimes lowered accuracy. The paper argues for selecting reasoning mode by task, effort and deployment context rather than enabling it universally.",
    "why_it_matters": "A 151-run analysis finds task structure predicts marginal reasoning efficiency better than nominal difficulty.",
    "limitations": [],
    "importance": 7,
    "canonical_url": "https://themachinepress.com/story/mp-2026-08-30-015/thinking-tokens-had-diminishing-returns",
    "json_url": "https://themachinepress.com/story/mp-2026-08-30-015.json",
    "first_published_at": "2026-08-30T09:00:00.000-04:00",
    "modified_at": "2026-08-30T09:00:00.000-04:00",
    "content_status": "new",
    "is_carryover": false,
    "carryover_reason": null,
    "key_claims": [
      {
        "claim_id": "claim-mp-2026-08-30-015-001",
        "text": "A 151-run analysis finds task structure predicts marginal reasoning efficiency better than nominal difficulty.",
        "source_ids": [
          "source-2026-08-30-017"
        ],
        "qualification": null
      }
    ],
    "source_ids": [
      "source-2026-08-30-017"
    ],
    "tags": [
      "reasoning models",
      "inference cost",
      "benchmarks"
    ],
    "image_url": null,
    "corrections": []
  },
  "sources": [
    {
      "source_id": "source-2026-08-30-017",
      "title": "arXiv preprint 2608.26235",
      "publisher": "arXiv",
      "url": "https://arxiv.org/abs/2608.26235",
      "canonical_url": "https://arxiv.org/abs/2608.26235",
      "source_type": "primary_research",
      "is_primary_source": true,
      "published_at": "2026-08-26T13:44:50.000-04:00",
      "accessed_at": "2026-08-30T08:23:14.000-04:00",
      "supports_claim_ids": [
        "claim-mp-2026-08-30-015-001"
      ]
    }
  ],
  "corrections": [],
  "publisher": {
    "name": "The Machine Press",
    "url": "https://themachinepress.com",
    "description": "A daily newspaper for the age of artificial intelligence."
  },
  "cite_this_report": {
    "title": "Thinking Tokens Had Diminishing Returns",
    "publisher": "The Machine Press",
    "published_at": "2026-08-30T09:00:00.000-04:00",
    "canonical_url": "https://themachinepress.com/story/mp-2026-08-30-015/thinking-tokens-had-diminishing-returns"
  }
}