{ "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": "2609.31291", "title": "Softmax Reparameterization for Output-Head Quantization", "summary": "It replaces direct post-training quantization of the output head (RTN, activation-weighted MSE, or GPTQ) with a functionally equivalent reparameterized head selected by a one-dimensional validation-KL search before quantization. The cost is offline calibration and search plus implementation effort; for shift-compatible heads it adds no inference operation and preserves packed W4 execution, while nonlinear logit paths such as soft-capping need a rank-one correction. It can fail when heads are not shift-compatible, when frozen coefficients do not transfer to a new data distribution, or when a given model-quantizer pair shows little baseline output-head distortion to recover.", "rationale": "It is a low-cost post-training change compatible with a single-GPU or API-constrained workflow, requires no retraining, and has a measured latency and KL benefit that the reader can validate on their own models.", "technique": "Softmax reparameterization for output-head quantization", "family": "quantizacao", "recommendation": "adotar", "published": "2026-09-25", "as_of": "2026-10-18", "editorial_status": "indexed", "source_url": "https://arxiv.org/abs/2609.31291", "independent_implementations": 0, "total_implementations": 0, "stars": 0, "citations": 0, "claims": [ { "claim_id": "claim-01", "statement": "It replaces direct post-training quantization of the output head (RTN, activation-weighted MSE, or GPTQ) with a functionally equivalent reparameterized head selected by a one-dimensional validation-KL search before quantization. The cost is offline calibration and search plus implementation effort; for shift-compatible heads it adds no inference operation and preserves packed W4 execution, while nonlinear logit paths such as soft-capping need a rank-one correction. It can fail when heads are not shift-compatible, when frozen coefficients do not transfer to a new data distribution, or when a given model-quantizer pair shows little baseline output-head distortion to recover.", "basis": "inferred", "result": "", "baseline": "", "conditions": "", "source_url": "https://arxiv.org/abs/2609.31291", "source_page": null, "source_excerpt": "" }, { "claim_id": "claim-02", "statement": "Quantizing the Phi output head to W4 with the decoder in BF16 reduces batch-one generation latency by 10.8% versus the BF16-head baseline; on Phi-4-mini, AW-MSE KL falls from 0.936 to 0.256.", "basis": "inferred", "result": "", "baseline": "", "conditions": "", "source_url": "https://arxiv.org/abs/2609.31291", "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": "" }