{ "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": "2610.27815", "title": "ProVe: Proposer-Verifier Decoding", "summary": "ProVe replaces the standard speculative-decoding arrangement (small serial drafter, large parallel verifier) by fine-tuning the large AR model to propose a block of future tokens in one forward pass, then verifying with a second pass from itself or a small model. It costs a fine-tuning run on the large checkpoint, one extra forward pass per block, and, in the lossy variant, a measurable accuracy drop (about 2.6 points on the reported six-task average); the lossless variant adds verification compute but preserves greedy AR outputs. It can fail if proposed blocks have short AR-matching prefixes on the target workload (gains depend on the 6.6-of-16 prefix statistic holding), if the fine-tune degrades base-model quality, or if self-verification overhead erodes the speedup at small batch sizes.", "rationale": "The method is a credible inference-time acceleration, but it requires fine-tuning and serving a 32B-class proposer, which exceeds a single 24 GB GPU and a limited cloud budget, so the reader should track it rather than adopt it now.", "technique": "Proposer-verifier decoding from a fine-tuned AR checkpoint", "family": "decodificacao_especulativa", "recommendation": "observar", "published": "2026-10-20", "as_of": "2026-10-21", "editorial_status": "indexed", "source_url": "https://arxiv.org/abs/2610.27815", "independent_implementations": 1, "total_implementations": 2, "stars": 122, "citations": null, "claims": [ { "claim_id": "claim-01", "statement": "ProVe replaces the standard speculative-decoding arrangement (small serial drafter, large parallel verifier) by fine-tuning the large AR model to propose a block of future tokens in one forward pass, then verifying with a second pass from itself or a small model. It costs a fine-tuning run on the large checkpoint, one extra forward pass per block, and, in the lossy variant, a measurable accuracy drop (about 2.6 points on the reported six-task average); the lossless variant adds verification compute but preserves greedy AR outputs. It can fail if proposed blocks have short AR-matching prefixes on the target workload (gains depend on the 6.6-of-16 prefix statistic holding), if the fine-tune degrades base-model quality, or if self-verification overhead erodes the speedup at small batch sizes.", "basis": "inferred", "result": "", "baseline": "", "conditions": "", "source_url": "https://arxiv.org/abs/2610.27815", "source_page": null, "source_excerpt": "" }, { "claim_id": "claim-02", "statement": "ProVe-lossless with a 32B proposer reaches 2.5x average AR throughput across six benchmarks without a separate draft head; with a 3B verifier it reaches 2.7x at 2.6 accuracy points below AR.", "basis": "inferred", "result": "", "baseline": "", "conditions": "", "source_url": "https://arxiv.org/abs/2610.27815", "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": "" }