{ "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": "2610.08164", "title": "Align, Then Correct: Training-Free Two-Stage Low-Rank Compensation for Extremely Quantized Large Language Models", "summary": "It replaces one-shot symmetric low-rank error compensation (e.g., standard LQEC adapters) with a two-stage closed-form procedure: a Fisher-weighted asymmetric alignment stage followed by a rank-constrained natural-gradient correction computed on re-measured statistics of the compensated model. Cost is one truncated SVD plus calibration forward and backward passes per layer for statistics, a small rank-r adapter added beside each frozen quantized weight (modest memory overhead on top of the quantized model), and additional compute during quantization only, not at serving. It can fail when calibration data is unrepresentative of deployment distribution, when the rank budget is too small for layers with high-rank residual error, or when gains do not transfer to quantizers or architectures beyond those evaluated.", "rationale": "The method is training-free, runs per-layer with closed-form SVD and statistics collection, and is directly applicable to the 2-4 bit weight quantization an engineer with a single 24 GB GPU already needs, with evidence on models of deployable size.", "technique": "Two-stage low-rank compensation for extreme quantization", "family": "quantizacao", "recommendation": "adotar", "published": "2026-10-06", "as_of": "2026-10-18", "editorial_status": "indexed", "source_url": "https://arxiv.org/abs/2610.08164", "independent_implementations": 0, "total_implementations": 0, "stars": 0, "citations": null, "claims": [ { "claim_id": "claim-01", "statement": "It replaces one-shot symmetric low-rank error compensation (e.g., standard LQEC adapters) with a two-stage closed-form procedure: a Fisher-weighted asymmetric alignment stage followed by a rank-constrained natural-gradient correction computed on re-measured statistics of the compensated model. Cost is one truncated SVD plus calibration forward and backward passes per layer for statistics, a small rank-r adapter added beside each frozen quantized weight (modest memory overhead on top of the quantized model), and additional compute during quantization only, not at serving. It can fail when calibration data is unrepresentative of deployment distribution, when the rank budget is too small for layers with high-rank residual error, or when gains do not transfer to quantizers or architectures beyond those evaluated.", "basis": "inferred", "result": "", "baseline": "", "conditions": "", "source_url": "https://arxiv.org/abs/2610.08164", "source_page": null, "source_excerpt": "" }, { "claim_id": "claim-02", "statement": "At 2 bits under QuIP#, WikiText-2 perplexity drops from 12.43 to 10.26 (Qwen3-8B) and 21.11 to 13.22 (Qwen3-4B); on held-out C4 it recovers 51% and 84% of the gap to FP16 versus 31% and 63% for the strongest baseline, with gains at higher bit-widths and under a second quantizer.", "basis": "inferred", "result": "", "baseline": "", "conditions": "", "source_url": "https://arxiv.org/abs/2610.08164", "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": "" }