{ "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": "2610.27123", "title": "Reconstruction-Optimized Expert Pruning for Mixture-of-Experts Language Models", "summary": "OptEP replaces heuristic or search-based expert-pruning criteria with a continuous relaxation that optimizes differentiable expert-selection variables against a reconstruction objective, followed by a lightweight update of retained expert weights; the pruned experts are removed, reducing storage and deployment memory for MoE models. Cost is a post-training calibration procedure requiring task-domain data, per-layer reconstruction optimization, and the weight-adaptation pass, plus some residual accuracy loss relative to the unpruned model. It can fail when calibration data is not representative of production traffic, when pruning is pushed to ratios the reconstruction objective cannot compensate for, and on hardware or serving stacks where sparse expert routing gives no realized speed or memory benefit despite fewer stored experts.", "rationale": "The method runs post-training on a single GPU with modest calibration data and directly reduces the storage footprint of MoE models the reader may already deploy, but the accuracy retention is workload- and architecture-dependent, so it warrants a small validation run before production adoption.", "technique": "Reconstruction-optimized MoE expert pruning (OptEP)", "family": "esparsidade_e_poda", "recommendation": "testar", "published": "2026-10-19", "as_of": "2026-10-25", "editorial_status": "indexed", "source_url": "https://arxiv.org/abs/2610.27123", "independent_implementations": 0, "total_implementations": 0, "stars": 0, "citations": 0, "claims": [ { "claim_id": "claim-01", "statement": "OptEP replaces heuristic or search-based expert-pruning criteria with a continuous relaxation that optimizes differentiable expert-selection variables against a reconstruction objective, followed by a lightweight update of retained expert weights; the pruned experts are removed, reducing storage and deployment memory for MoE models. Cost is a post-training calibration procedure requiring task-domain data, per-layer reconstruction optimization, and the weight-adaptation pass, plus some residual accuracy loss relative to the unpruned model. It can fail when calibration data is not representative of production traffic, when pruning is pushed to ratios the reconstruction objective cannot compensate for, and on hardware or serving stacks where sparse expert routing gives no realized speed or memory benefit despite fewer stored experts.", "basis": "inferred", "result": "", "baseline": "", "conditions": "", "source_url": "https://arxiv.org/abs/2610.27123", "source_page": null, "source_excerpt": "" }, { "claim_id": "claim-02", "statement": "Absolute accuracy gains of up to 15.05% on Qwen3-30B-A3B, 14.55% on Qwen3-235B-A22B, 11.06% on GLM-4.5-Air, and 16.20% on Mixtral-8x7B over the strongest expert-pruning baselines, with the largest improvements under aggressive pruning.", "basis": "inferred", "result": "", "baseline": "", "conditions": "", "source_url": "https://arxiv.org/abs/2610.27123", "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": "" }