{ "schema_version": 1, "builder_plan": { "scenario": "A self-hosted model uses more compute or memory than your workload can justify.", "change": "Check for a usable pruned checkpoint and runtime support before attempting model surgery.", "baseline": "The original unpruned model at the same precision, context, and load.", "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 quality, actual GPU memory, serving throughput, and preparation 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": "2609.29032", "title": "Paging the Experts: A Reproducible Characterization of Flash-Backed MoE Inference on iPhone", "summary": "The method replaces fully RAM-resident expert weights in MoE inference with a byte-budgeted in-memory expert cache paged from flash storage, targeting mobile hardware. The cost is strong sensitivity of cache hit rate to eviction policy and workload (0.00% to 38.58% at comparable budgets), plus prefetch machinery and runtime complexity; the paper reports no latency advantage and retains a thermal stopping event and negative timing comparisons. What can fail after adoption is numerical equivalence (resident-Python versus recorded-phone sequences disagreed on all five tested cases), cross-device reproducibility, and thermal or timing behavior under sustained load.", "rationale": "The technique addresses iPhone-class edge deployment with a Swift/MLX runtime and is explicitly framed as bounded feasibility rather than a deployment-ready result, so it does not transfer to the reader's 24 GB GPU or API-based production setting.", "technique": "Flash-backed expert paging for on-device MoE inference", "family": "esparsidade_e_poda", "recommendation": "observar", "published": "2026-09-24", "as_of": "2026-10-18", "editorial_status": "indexed", "source_url": "https://arxiv.org/abs/2609.29032", "independent_implementations": 0, "total_implementations": 1, "stars": 0, "citations": 0, "claims": [ { "claim_id": "claim-01", "statement": "The method replaces fully RAM-resident expert weights in MoE inference with a byte-budgeted in-memory expert cache paged from flash storage, targeting mobile hardware. The cost is strong sensitivity of cache hit rate to eviction policy and workload (0.00% to 38.58% at comparable budgets), plus prefetch machinery and runtime complexity; the paper reports no latency advantage and retains a thermal stopping event and negative timing comparisons. What can fail after adoption is numerical equivalence (resident-Python versus recorded-phone sequences disagreed on all five tested cases), cross-device reproducibility, and thermal or timing behavior under sustained load.", "basis": "inferred", "result": "", "baseline": "", "conditions": "", "source_url": "https://arxiv.org/abs/2609.29032", "source_page": null, "source_excerpt": "" }, { "claim_id": "claim-02", "statement": "Runs the text path of a quantized Qwen3.6-35B-A3B MoE on iPhone with expert weights in flash storage and sampled process-footprint peaks of 1.87-2.73 GiB; cache hit rates ranged from 0.00% (512 MiB LRU, fixed-route replay) to 38.58% (576 MiB LRU), showing feasibility is a policy/workload interaction rather than a fixed capacity requirement.", "basis": "inferred", "result": "", "baseline": "", "conditions": "", "source_url": "https://arxiv.org/abs/2609.29032", "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": "" }