{
  "$schema": "https://themachinepress.com/schemas/story-v1.schema.json",
  "schema_version": "1.0.0",
  "document_type": "machine_press_story",
  "story": {
    "story_id": "mp-2026-10-31-005",
    "source_story_id": "tmp-story-decode-length-prediction-limits",
    "edition_id": "mp-2026-10-31-morning-0113",
    "edition_url": "https://themachinepress.com/edition/2026-10-31",
    "position": 5,
    "story_type": "dispatch",
    "section": "infrastructure",
    "editorial_classification": "editorial",
    "headline": "The Scheduler Cannot Predict Away Sampling Noise",
    "slug": "the-scheduler-cannot-predict-away-sampling-noise",
    "dek": "Prefill-side decode-length predictors converged near the same bound once model, data and sampling conditions were fixed.",
    "summary": "Prefill-side decode-length predictors converged near the same bound once model, data and sampling conditions were fixed.",
    "body_text": "Inference schedulers would benefit from knowing how long a response will be before generation starts. This analysis treats decoding as a stopped stochastic process and argues that autoregressive sampling places a floor under any predictor that sees only prefill hidden states. Across several predictor designs, results under fixed model, dataset and sampling settings converged near the derived bound, suggesting system randomness dominated model-design differences. The authors propose benchmarking with fixed generation conditions and relative-error metrics; the paper derives and tests limits rather than a universal production scheduler.",
    "why_it_matters": "Prefill-side decode-length predictors converged near the same bound once model, data and sampling conditions were fixed.",
    "limitations": [
      "This analysis treats decoding as a stopped stochastic process and argues that autoregressive sampling places a floor under any predictor that sees only prefill hidden states."
    ],
    "importance": 8,
    "canonical_url": "https://themachinepress.com/story/mp-2026-10-31-005/the-scheduler-cannot-predict-away-sampling-noise",
    "json_url": "https://themachinepress.com/story/mp-2026-10-31-005.json",
    "first_published_at": "2026-10-31T09:00:00.000-04:00",
    "modified_at": "2026-10-31T09:00:00.000-04:00",
    "content_status": "new",
    "is_carryover": false,
    "carryover_reason": null,
    "key_claims": [
      {
        "claim_id": "claim-mp-2026-10-31-005-001",
        "text": "Prefill-side decode-length predictors converged near the same bound once model, data and sampling conditions were fixed.",
        "source_ids": [
          "source-2026-10-31-005"
        ],
        "qualification": "This analysis treats decoding as a stopped stochastic process and argues that autoregressive sampling places a floor under any predictor that sees only prefill hidden states."
      }
    ],
    "source_ids": [
      "source-2026-10-31-005"
    ],
    "tags": [
      "LLM inference",
      "scheduling",
      "sampling",
      "benchmarks"
    ],
    "image_url": null,
    "corrections": []
  },
  "sources": [
    {
      "source_id": "source-2026-10-31-005",
      "title": "arXiv preprint 2610.40541, original v1",
      "publisher": "arXiv",
      "url": "https://arxiv.org/abs/2610.40541v1",
      "canonical_url": "https://arxiv.org/abs/2610.40541v1",
      "source_type": "primary_research",
      "is_primary_source": true,
      "published_at": "2026-10-28T20:00:00.000-04:00",
      "accessed_at": "2026-10-31T08:25:01.279-04:00",
      "supports_claim_ids": [
        "claim-mp-2026-10-31-005-001"
      ]
    }
  ],
  "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 Scheduler Cannot Predict Away Sampling Noise",
    "publisher": "The Machine Press",
    "published_at": "2026-10-31T09:00:00.000-04:00",
    "canonical_url": "https://themachinepress.com/story/mp-2026-10-31-005/the-scheduler-cannot-predict-away-sampling-noise"
  }
}