{
  "$schema": "https://themachinepress.com/schemas/story-v1.schema.json",
  "schema_version": "1.0.0",
  "document_type": "machine_press_story",
  "story": {
    "story_id": "mp-2026-10-14-007",
    "source_story_id": "tmp-story-doubtless-contrastive-decoding",
    "edition_id": "mp-2026-10-14-morning-0096",
    "edition_url": "https://themachinepress.com/edition/2026-10-14",
    "position": 7,
    "story_type": "dispatch",
    "section": "frontier-models",
    "editorial_classification": "editorial",
    "headline": "A Smaller Model Tells the Larger One When to Stop Rechecking",
    "slug": "a-smaller-model-tells-the-larger-one-when-to-stop-rechecking",
    "dek": "DoubtLess uses contrasting brief and verbose predictions from a small model to steer reasoning without retraining the large one.",
    "summary": "DoubtLess uses contrasting brief and verbose predictions from a small model to steer reasoning without retraining the large one.",
    "body_text": "DoubtLess reads a large model's reasoning trace with a smaller model twice: once under a brief instruction and once under a verbose one. The difference between those token predictions nudges the larger model during decoding. Across seven math, science and programming benchmarks, the authors report cutting Qwen3-32B reasoning length by 31 percent without changing accuracy; a self-debugging coding agent used 20 percent fewer thinking tokens per turn. The method still adds a second model at inference and its unchanged accuracy is measured on the selected tasks. It offers a training-free route to reducing repeated checking after a correct answer has already emerged.",
    "why_it_matters": "DoubtLess uses contrasting brief and verbose predictions from a small model to steer reasoning without retraining the large one.",
    "limitations": [],
    "importance": 7,
    "canonical_url": "https://themachinepress.com/story/mp-2026-10-14-007/a-smaller-model-tells-the-larger-one-when-to-stop-rechecking",
    "json_url": "https://themachinepress.com/story/mp-2026-10-14-007.json",
    "first_published_at": "2026-10-14T09:00:00.000-04:00",
    "modified_at": "2026-10-14T09:00:00.000-04:00",
    "content_status": "new",
    "is_carryover": false,
    "carryover_reason": null,
    "key_claims": [
      {
        "claim_id": "claim-mp-2026-10-14-007-001",
        "text": "DoubtLess uses contrasting brief and verbose predictions from a small model to steer reasoning without retraining the large one.",
        "source_ids": [
          "source-2026-10-14-007"
        ],
        "qualification": null
      }
    ],
    "source_ids": [
      "source-2026-10-14-007"
    ],
    "tags": [
      "DoubtLess",
      "reasoning efficiency",
      "contrastive decoding"
    ],
    "image_url": null,
    "corrections": []
  },
  "sources": [
    {
      "source_id": "source-2026-10-14-007",
      "title": "arXiv preprint 2610.18862, original v1",
      "publisher": "arXiv / submitting research authors",
      "url": "https://arxiv.org/abs/2610.18862v1",
      "canonical_url": "https://arxiv.org/abs/2610.18862v1",
      "source_type": "primary_research",
      "is_primary_source": true,
      "published_at": "2026-10-13T13:59:32.000-04:00",
      "accessed_at": "2026-10-14T08:24:30.364-04:00",
      "supports_claim_ids": [
        "claim-mp-2026-10-14-007-001"
      ]
    }
  ],
  "corrections": [],
  "publisher": {
    "name": "The Machine Press",
    "url": "https://themachinepress.com",
    "description": "A daily newspaper for the age of artificial intelligence."
  },
  "cite_this_report": {
    "title": "A Smaller Model Tells the Larger One When to Stop Rechecking",
    "publisher": "The Machine Press",
    "published_at": "2026-10-14T09:00:00.000-04:00",
    "canonical_url": "https://themachinepress.com/story/mp-2026-10-14-007/a-smaller-model-tells-the-larger-one-when-to-stop-rechecking"
  }
}