{ "schema_version": 1, "builder_plan": { "scenario": "A self-hosted assistant is too expensive because its model uses most of the GPU memory.", "change": "Try the paper's supported lower-precision weights on a copy of your serving setup.", "baseline": "The same model at its original precision, on the same GPU.", "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": "Peak memory, task pass rate, latency, and total serving cost at the same context and load.", "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": "2610.09877", "title": "Layerwise Error Attribution for Fast and Robust Mixed-Precision Post-Training Quantization", "summary": "Replaces solver-based or search-based mixed-precision bit-allocation methods in post-training quantization with a separable layerwise score derived from a probabilistic decomposition of propagated versus local quantization error, computed without external solvers. Costs are low: a small calibration set and a fast scoring pass, with allocation overhead reduced by orders of magnitude, though the method still requires per-layer sensitivity computation and calibration data. Can fail when the probabilistic layerwise error assumptions do not hold for a given architecture, and its evidence base is denoising networks (DRUNet) and diffusion models, so gains on LLMs or other production workloads are not established.", "rationale": "The method is cheap and solver-free, which suits constrained infrastructure, but its demonstrated results are limited to image denoising and diffusion models, so adoption for other model families requires workload-specific validation.", "technique": "Layerwise error-attribution mixed-precision PTQ bit allocation", "family": "quantizacao", "recommendation": "testar", "published": "2026-10-07", "as_of": "2026-10-18", "editorial_status": "indexed", "source_url": "https://arxiv.org/abs/2610.09877", "independent_implementations": 1, "total_implementations": 2, "stars": 4, "citations": null, "claims": [ { "claim_id": "claim-01", "statement": "Replaces solver-based or search-based mixed-precision bit-allocation methods in post-training quantization with a separable layerwise score derived from a probabilistic decomposition of propagated versus local quantization error, computed without external solvers. Costs are low: a small calibration set and a fast scoring pass, with allocation overhead reduced by orders of magnitude, though the method still requires per-layer sensitivity computation and calibration data. Can fail when the probabilistic layerwise error assumptions do not hold for a given architecture, and its evidence base is denoising networks (DRUNet) and diffusion models, so gains on LLMs or other production workloads are not established.", "basis": "inferred", "result": "", "baseline": "", "conditions": "", "source_url": "https://arxiv.org/abs/2610.09877", "source_page": null, "source_excerpt": "" }, { "claim_id": "claim-02", "statement": "Bit-allocation speed-ups from 28x to 2,570x over mixed-precision baselines; PSNR gains up to 7.5 dB under corrupted calibration at 4-bit average on DRUNet denoising, with improvements on quantized diffusion models.", "basis": "inferred", "result": "", "baseline": "", "conditions": "", "source_url": "https://arxiv.org/abs/2610.09877", "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": "" }