{ "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": "2610.21695", "title": "ShiftQAT: Quantization-Aware Training for the Numerical Contract of a Fixed-Function Edge CNN Accelerator", "summary": "ShiftQAT replaces generic post-training quantization or standard W8A8 QAT with a hardware-aware QAT flow that enforces power-of-two per-tensor scales, integer-only arithmetic, and one static requantization shift per layer, so models satisfy the MAX7800x numerical contract by construction. The cost is coupling to the ai8x-training toolchain and the contract's constraints (signed 8-bit activation storage, 8-bit bias budget), plus up to 2.9 percentage points of accuracy loss at W4A8 on ImageNet. Adoption fails if the target deployment is not the MAX7800x family, if the model uses operators the fixed-function accelerator cannot express, or if scale-propagation constraints force architecture changes that degrade task accuracy.", "rationale": "The method is only actionable for teams deploying CNNs on MAX7800x-class fixed-function accelerators, which the reader's GPU/API-centric infrastructure does not include.", "technique": "QAT for fixed-function edge CNN accelerator contract", "family": "quantizacao", "recommendation": "observar", "published": "2026-10-15", "as_of": "2026-10-18", "editorial_status": "indexed", "source_url": "https://arxiv.org/abs/2610.21695", "independent_implementations": 0, "total_implementations": 0, "stars": 0, "citations": null, "claims": [ { "claim_id": "claim-01", "statement": "ShiftQAT replaces generic post-training quantization or standard W8A8 QAT with a hardware-aware QAT flow that enforces power-of-two per-tensor scales, integer-only arithmetic, and one static requantization shift per layer, so models satisfy the MAX7800x numerical contract by construction. The cost is coupling to the ai8x-training toolchain and the contract's constraints (signed 8-bit activation storage, 8-bit bias budget), plus up to 2.9 percentage points of accuracy loss at W4A8 on ImageNet. Adoption fails if the target deployment is not the MAX7800x family, if the model uses operators the fixed-function accelerator cannot express, or if scale-propagation constraints force architecture changes that degrade task accuracy.", "basis": "inferred", "result": "", "baseline": "", "conditions": "", "source_url": "https://arxiv.org/abs/2610.21695", "source_page": null, "source_excerpt": "" }, { "claim_id": "claim-02", "statement": "On MAX78002 silicon, ShiftQAT-trained MobileNetV2 and ResNet-18 classify ImageNet images in 14.6 and 8.4 ms at 5.1 and 2.8 mJ per inference; W8A8 accuracy matches less constrained QAT methods and W4A8 stays within 2.9 points of FP32.", "basis": "inferred", "result": "", "baseline": "", "conditions": "", "source_url": "https://arxiv.org/abs/2610.21695", "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": "" }