{
  "$schema": "https://themachinepress.com/schemas/story-v1.schema.json",
  "schema_version": "1.0.0",
  "document_type": "machine_press_story",
  "story": {
    "story_id": "mp-2026-10-06-006",
    "source_story_id": "tmp-story-budgetapo-tight-budget",
    "edition_id": "mp-2026-10-06-morning-0089",
    "edition_url": "https://themachinepress.com/edition/2026-10-06",
    "position": 6,
    "story_type": "dispatch",
    "section": "benchmarks-evals",
    "editorial_classification": "editorial",
    "headline": "A Prompt Optimizer Learned How Much Evidence It Could Afford",
    "slug": "a-prompt-optimizer-learned-how-much-evidence-it-could-afford",
    "dek": "BudgetAPO sized evaluation slices to task noise and kept every comparison paired under strict call limits.",
    "summary": "BudgetAPO sized evaluation slices to task noise and kept every comparison paired under strict call limits.",
    "body_text": "Automatic prompt optimizers often assume hundreds or thousands of model calls, a poor fit for paid or rate-limited systems. BudgetAPO begins with a short noise probe, sizes each evaluation slice to the task, and compares candidates on a fixed paired slice. Across seven benchmarks and five subject models, the authors report that it ranked first for every subject under their tests. At a 250-call budget it returned the unchanged seed prompt in 13% of runs, versus 86% for GEPA; on GPT-OSS-20B, GEPA needed 4.5 times as many calls to match BudgetAPO's 100-call score. The result turns budget allocation into part of the optimizer rather than an external limit it discovers too late.",
    "why_it_matters": "BudgetAPO sized evaluation slices to task noise and kept every comparison paired under strict call limits.",
    "limitations": [],
    "importance": 7,
    "canonical_url": "https://themachinepress.com/story/mp-2026-10-06-006/a-prompt-optimizer-learned-how-much-evidence-it-could-afford",
    "json_url": "https://themachinepress.com/story/mp-2026-10-06-006.json",
    "first_published_at": "2026-10-06T09:00:00.000-04:00",
    "modified_at": "2026-10-06T09:00:00.000-04:00",
    "content_status": "new",
    "is_carryover": false,
    "carryover_reason": null,
    "key_claims": [
      {
        "claim_id": "claim-mp-2026-10-06-006-001",
        "text": "BudgetAPO sized evaluation slices to task noise and kept every comparison paired under strict call limits.",
        "source_ids": [
          "source-2026-10-06-006"
        ],
        "qualification": null
      }
    ],
    "source_ids": [
      "source-2026-10-06-006"
    ],
    "tags": [
      "prompt optimization",
      "evaluation budget",
      "LLM operations"
    ],
    "image_url": null,
    "corrections": []
  },
  "sources": [
    {
      "source_id": "source-2026-10-06-006",
      "title": "arXiv preprint 2610.05671",
      "publisher": "arXiv",
      "url": "https://arxiv.org/abs/2610.05671",
      "canonical_url": "https://arxiv.org/abs/2610.05671",
      "source_type": "primary_research",
      "is_primary_source": true,
      "published_at": "2026-10-04T21:34:14.000-04:00",
      "accessed_at": "2026-10-06T08:25:42.165-04:00",
      "supports_claim_ids": [
        "claim-mp-2026-10-06-006-001"
      ]
    }
  ],
  "corrections": [],
  "publisher": {
    "name": "The Machine Press",
    "url": "https://themachinepress.com",
    "description": "A daily newspaper for the age of artificial intelligence."
  },
  "cite_this_report": {
    "title": "A Prompt Optimizer Learned How Much Evidence It Could Afford",
    "publisher": "The Machine Press",
    "published_at": "2026-10-06T09:00:00.000-04:00",
    "canonical_url": "https://themachinepress.com/story/mp-2026-10-06-006/a-prompt-optimizer-learned-how-much-evidence-it-could-afford"
  }
}