{ "schema_version": 1, "builder_plan": { "scenario": "Your workload might fit a smaller or differently structured model.", "change": "Check for available weights and a compatible runtime before planning a product trial.", "baseline": "Your current model on identical held-out tasks and the same resource budget.", "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": "Task success, latency, memory, context limits, and migration cost.", "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": "2609.16367", "title": "FINNAS: FINN-Guided Hardware-Aware NAS and Pruning for FPGA Jet Substructure Classification", "summary": "The method replaces manual design-space exploration of quantised MLP accelerators for FPGAs with an evolutionary NAS that jointly searches depth, width, and precision, ranking candidates with FINN-estimated LUT and latency proxies before pruning and RTL validation. It costs access to the FINN/Vivado FPGA toolchain, repeated synthesis-in-the-loop evaluation during search, and acceptance of fully parallel mappings whose proxy estimates must be confirmed by out-of-context synthesis. It can fail when the search proxies mispredict post-synthesis resource or latency figures, when results do not transfer off the specific jet-substructure task and target FPGA, and it is irrelevant to deployments on GPU or API-served models.", "rationale": "The technique targets FPGA accelerator design, which the reader's single-GPU/API infrastructure does not involve, so it is only worth tracking in case future edge-hardware deployment becomes a requirement.", "technique": "Hardware-aware evolutionary NAS for FPGA QNNs", "family": "arquitetura_eficiente", "recommendation": "observar", "published": "2026-09-14", "as_of": "2026-10-18", "editorial_status": "indexed", "source_url": "https://arxiv.org/abs/2609.16367", "independent_implementations": 0, "total_implementations": 0, "stars": 0, "citations": 0, "claims": [ { "claim_id": "claim-01", "statement": "The method replaces manual design-space exploration of quantised MLP accelerators for FPGAs with an evolutionary NAS that jointly searches depth, width, and precision, ranking candidates with FINN-estimated LUT and latency proxies before pruning and RTL validation. It costs access to the FINN/Vivado FPGA toolchain, repeated synthesis-in-the-loop evaluation during search, and acceptance of fully parallel mappings whose proxy estimates must be confirmed by out-of-context synthesis. It can fail when the search proxies mispredict post-synthesis resource or latency figures, when results do not transfer off the specific jet-substructure task and target FPGA, and it is irrelevant to deployments on GPU or API-served models.", "basis": "inferred", "result": "", "baseline": "", "conditions": "", "source_url": "https://arxiv.org/abs/2609.16367", "source_page": null, "source_excerpt": "" }, { "claim_id": "claim-02", "statement": "A searched compact design improves accuracy from 73.78% to 74.36% while reducing LUT usage by 8.5x and RTL-simulation latency by 1.77x versus a manually optimised dense FINN accelerator; unstructured pruning adds further LUT and FF reductions.", "basis": "inferred", "result": "", "baseline": "", "conditions": "", "source_url": "https://arxiv.org/abs/2609.16367", "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": [ { "anchor": "S2.E2", "label": "(2)", "section": "II Methodology and Design", "role": "proposed_method", "latex": "R(x)=A(x)\\cdot\\max\\left(1,\\frac{L(x)}{B_{\\mathrm{l}}}\\right)^{-\\beta_{\\mathrm{l}}}\\cdot\\max\\left(1,\\frac{T(x)}{B_{\\mathrm{t}}}\\right)^{-\\beta_{\\mathrm{t}}}", "mathml": "R⁡(x)=A⁡(x)⋅max⁡(1,L⁡(x)Bl)−βl⋅max⁡(1,T⁡(x)Bt)−βt", "context": "The estimated accuracy and hardware metrics are combined through the reward R⁡(x) defined in Equation 2.", "headline": "", "explanation": "", "objective": "", "symbols": {}, "evidence": "", "evidence_url": "", "source_url": "" }, { "anchor": "S2.E1", "label": "(1)", "section": "II Methodology and Design", "role": "proposed_method", "latex": "x=\\{[h_{1},\\ldots,h_{L}],(B_{w},B_{ia},B_{ha},B_{oa})\\}", "mathml": "x={[h1,…,hL],(Bw,Bi​a,Bh​a,Bo​a)}", "context": "Each individual is encoded as Equation 1, i.e., an ordered list of hidden-layer widths and four global precision fields. At initialisation, L, every hi, and each precision field are drawn from their corresponding finite sets in Table I. Samples are repaired until they satisfy Bo​a≥Bh​a and Bw<8 when all activation precisions are below 8 bits. Each candidate then undergoes 16 epochs of proxy training using cross-entropy loss and the AdamW optimiser to obtain validation accuracy A⁡(x).", "headline": "", "explanation": "", "objective": "", "symbols": {}, "evidence": "", "evidence_url": "", "source_url": "" } ], "equations_status": "selected", "core_kind": "formula", "equations_fetched_at": "2026-10-13", "equations_source_url": "", "equations_source_sha256": "", "exposures_reviewed_at": "", "exposures_source_sha256": "" }