{
  "schema_version": 1,
  "builder_plan": {
    "scenario": "Your self-hosted model spends too long generating answers token by token.",
    "change": "Check whether a compatible draft model and serving integration exist before renting compute.",
    "baseline": "The same target with speculation off, then your existing drafter if you have one.",
    "steps": [
      "Locate the paper's implementation and check its license, model support, data needs, and hardware requirements. Stop here if you cannot obtain the required artifacts.",
      "Save a small set of representative product tasks and expected outcomes. Keep evaluation tasks out of any training or tuning, and choose quality and cost limits before running.",
      "Run the baseline and the proposed change on identical inputs. Keep other settings fixed and repeat paired runs to expose variation.",
      "Save per-task outputs, failures, timings, and all usage costs. Review regressions before expanding the test."
    ],
    "measure": "Accepted draft length, tokens per second, tail latency, output quality, and total GPU cost.",
    "decision": "Keep the change only if it meets your preselected quality and latency limits and improves the metric you care about after setup and operating costs. Otherwise retain the baseline. This is an exploration plan, not a validated recommendation."
  },
  "builder_review": null,
  "arxiv_id": "2609.17943",
  "title": "ASPIRE: Asynchronous Batched Self-Speculative Decoding for Long-Context LLM Inference",
  "summary": "ASPIRE replaces synchronized draft-verify scheduling in batched self-speculative decoding with a mixed forward pass where drafting and verifying requests coexist, plus per-request acceptance-rate scheduling and a single full-attention refresh layer during drafting. It costs implementation complexity in the serving stack: a custom unified kernel path, an online scheduler with a batch-aware cost model, and modified drafting with an intra-draft refresh layer, none of which are available in standard inference frameworks. It can fail if per-request acceptance estimates are noisy or unstable, if the cost model misjudges batch composition, or if sparse-attention drafting quality degrades on workloads unlike the evaluated reasoning and long-context benchmarks.",
  "rationale": "The gains are substantial but realizing them requires modifying the inference engine's batching and attention internals, which is beyond what a small team on a single GPU or third-party APIs can adopt until integrated into a maintained serving framework.",
  "technique": "Asynchronous batched self-speculative decoding",
  "family": "decodificacao_especulativa",
  "recommendation": "observar",
  "published": "2026-09-16",
  "as_of": "2026-10-18",
  "editorial_status": "indexed",
  "source_url": "https://arxiv.org/abs/2609.17943",
  "independent_implementations": 0,
  "total_implementations": 0,
  "stars": 0,
  "citations": null,
  "claims": [
    {
      "claim_id": "claim-01",
      "statement": "ASPIRE replaces synchronized draft-verify scheduling in batched self-speculative decoding with a mixed forward pass where drafting and verifying requests coexist, plus per-request acceptance-rate scheduling and a single full-attention refresh layer during drafting. It costs implementation complexity in the serving stack: a custom unified kernel path, an online scheduler with a batch-aware cost model, and modified drafting with an intra-draft refresh layer, none of which are available in standard inference frameworks. It can fail if per-request acceptance estimates are noisy or unstable, if the cost model misjudges batch composition, or if sparse-attention drafting quality degrades on workloads unlike the evaluated reasoning and long-context benchmarks.",
      "basis": "inferred",
      "result": "",
      "baseline": "",
      "conditions": "",
      "source_url": "https://arxiv.org/abs/2609.17943",
      "source_page": null,
      "source_excerpt": ""
    },
    {
      "claim_id": "claim-02",
      "statement": "1.70-4.58x decoding throughput speedup over autoregressive baselines across three models and five benchmarks, and approximately 27% improvement over the strongest prior self-speculative baseline.",
      "basis": "inferred",
      "result": "",
      "baseline": "",
      "conditions": "",
      "source_url": "https://arxiv.org/abs/2609.17943",
      "source_page": null,
      "source_excerpt": ""
    }
  ],
  "exposure_map": [
    {
      "dimension": "quality",
      "basis": "not_evaluated",
      "finding": "",
      "interpretation": "",
      "limitation": "",
      "next_check": "",
      "source_url": "",
      "source_page": null,
      "source_excerpt": ""
    },
    {
      "dimension": "compute",
      "basis": "not_evaluated",
      "finding": "",
      "interpretation": "",
      "limitation": "",
      "next_check": "",
      "source_url": "",
      "source_page": null,
      "source_excerpt": ""
    },
    {
      "dimension": "latency",
      "basis": "not_evaluated",
      "finding": "",
      "interpretation": "",
      "limitation": "",
      "next_check": "",
      "source_url": "",
      "source_page": null,
      "source_excerpt": ""
    },
    {
      "dimension": "operations",
      "basis": "not_evaluated",
      "finding": "",
      "interpretation": "",
      "limitation": "",
      "next_check": "",
      "source_url": "",
      "source_page": null,
      "source_excerpt": ""
    },
    {
      "dimension": "compatibility",
      "basis": "not_evaluated",
      "finding": "",
      "interpretation": "",
      "limitation": "",
      "next_check": "",
      "source_url": "",
      "source_page": null,
      "source_excerpt": ""
    },
    {
      "dimension": "security",
      "basis": "not_evaluated",
      "finding": "",
      "interpretation": "",
      "limitation": "",
      "next_check": "",
      "source_url": "",
      "source_page": null,
      "source_excerpt": ""
    },
    {
      "dimension": "data_and_training",
      "basis": "not_evaluated",
      "finding": "",
      "interpretation": "",
      "limitation": "",
      "next_check": "",
      "source_url": "",
      "source_page": null,
      "source_excerpt": ""
    },
    {
      "dimension": "reproducibility",
      "basis": "not_evaluated",
      "finding": "",
      "interpretation": "",
      "limitation": "",
      "next_check": "",
      "source_url": "",
      "source_page": null,
      "source_excerpt": ""
    }
  ],
  "risks": [],
  "minimum_test": [],
  "open_questions": [],
  "independent_tests": [],
  "lab_reproductions": [],
  "report_available": false,
  "validation_tier": "unknown",
  "evidence_tier": "unknown",
  "infrastructure_basis": "unknown",
  "equations": [],
  "equations_status": "not_fetched",
  "core_kind": "",
  "equations_fetched_at": "",
  "equations_source_url": "",
  "equations_source_sha256": "",
  "exposures_reviewed_at": "",
  "exposures_source_sha256": ""
}