{ "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": "2610.09342", "title": "Shared Low-rank Basis Factorization for Data-free Mixture-of-Experts Compression", "summary": "Replaces expert pruning and expert merging for reducing MoE model size, instead reconstructing each expert's weights from a rank-k basis shared across experts, plus a post-hoc gauge fixing step that removes redundant parameters. It is data-free, so it avoids calibration data and retraining, but it costs a reconstruction pass over the full model and adds factorization/gauge-fixing implementation complexity; the compressed model also requires kernels or formats that exploit the low-rank shared structure to realize storage or serving gains. It can fail if the chosen rank is too low for heterogeneous experts, if reconstruction error compounds in downstream task accuracy, or if the reader's serving stack cannot exploit the factored weights, and applying it to 16B–122B models requires memory headroom beyond a single 24 GB GPU during compression.", "rationale": "The method is credible and data-free, but it targets large MoE checkpoints whose compression and validation exceed the reader's single-GPU budget, so it is worth tracking rather than adopting now.", "technique": "Shared low-rank basis factorization for MoE compression", "family": "esparsidade_e_poda", "recommendation": "observar", "published": "2026-10-07", "as_of": "2026-10-18", "editorial_status": "indexed", "source_url": "https://arxiv.org/abs/2610.09342", "independent_implementations": 0, "total_implementations": 0, "stars": 0, "citations": 0, "claims": [ { "claim_id": "claim-01", "statement": "Replaces expert pruning and expert merging for reducing MoE model size, instead reconstructing each expert's weights from a rank-k basis shared across experts, plus a post-hoc gauge fixing step that removes redundant parameters. It is data-free, so it avoids calibration data and retraining, but it costs a reconstruction pass over the full model and adds factorization/gauge-fixing implementation complexity; the compressed model also requires kernels or formats that exploit the low-rank shared structure to realize storage or serving gains. It can fail if the chosen rank is too low for heterogeneous experts, if reconstruction error compounds in downstream task accuracy, or if the reader's serving stack cannot exploit the factored weights, and applying it to 16B–122B models requires memory headroom beyond a single 24 GB GPU during compression.", "basis": "inferred", "result": "", "baseline": "", "conditions": "", "source_url": "https://arxiv.org/abs/2610.09342", "source_page": null, "source_excerpt": "" }, { "claim_id": "claim-02", "statement": "SLBF consistently outperforms pruning, merging, and other reconstruction methods across five MoE architectures (16B to 122B parameters), with lower reconstruction error; no multiplicative factor is reported.", "basis": "inferred", "result": "", "baseline": "", "conditions": "", "source_url": "https://arxiv.org/abs/2610.09342", "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": "" }