{ "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": "2610.00983", "title": "The Devil Is in the Reconstruction Loss Scale: Rethinking Optimization in LLM Quantization", "summary": "The technique replaces mean squared error with root mean squared error variants as the reconstruction loss when optimizing auxiliary quantization parameters in sequential learning-based PTQ. It costs nothing extra in memory or latency, since the square root is applied to the loss during the same gradient-based optimization already performed. It can fail if a given PTQ implementation already compensates for loss-scale imbalance through tuned per-stage learning rates, in which case the improvement may be smaller than reported, and the cross-stage loss-scale diagnosis should be verified on the target model before assuming benefit.", "rationale": "It is a zero-cost, drop-in loss change applicable to PTQ workflows that a 24 GB GPU can already run, with consistent quality gains reported across methods and models.", "technique": "RMSE reconstruction loss for learning-based PTQ", "family": "quantizacao", "recommendation": "adotar", "published": "2026-10-01", "as_of": "2026-10-18", "editorial_status": "indexed", "source_url": "https://arxiv.org/abs/2610.00983", "independent_implementations": 1, "total_implementations": 2, "stars": 125, "citations": null, "claims": [ { "claim_id": "claim-01", "statement": "The technique replaces mean squared error with root mean squared error variants as the reconstruction loss when optimizing auxiliary quantization parameters in sequential learning-based PTQ. It costs nothing extra in memory or latency, since the square root is applied to the loss during the same gradient-based optimization already performed. It can fail if a given PTQ implementation already compensates for loss-scale imbalance through tuned per-stage learning rates, in which case the improvement may be smaller than reported, and the cross-stage loss-scale diagnosis should be verified on the target model before assuming benefit.", "basis": "inferred", "result": "", "baseline": "", "conditions": "", "source_url": "https://arxiv.org/abs/2610.00983", "source_page": null, "source_excerpt": "" }, { "claim_id": "claim-02", "statement": "The paper reports that RMSE loss variants (sample, channel, token, and element level) significantly outperform MSE as a drop-in replacement across representative PTQ methods, model families, scales, and quantization settings, without providing a single multiplicative factor.", "basis": "inferred", "result": "", "baseline": "", "conditions": "", "source_url": "https://arxiv.org/abs/2610.00983", "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": "" }