{ "schema_version": 1, "builder_plan": { "scenario": "You already need task-specific training and want to reduce its memory or compute cost.", "change": "Change only the training technique on a small controlled run after checking prerequisites.", "baseline": "Your current training recipe with the same model, data split, and evaluation.", "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": "Held-out quality, peak memory, training time, and full compute cost to the target quality.", "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": "2610.21201", "title": "Hard Way, The Only Way: Provable Merits of Training on Hardest Examples", "summary": "Replaces training on the full dataset with drastic pruning that retains only the hardest examples (as few as 2N) under weighted empirical risk minimization. Costs are a scoring/ranking pass over the training set to identify hard examples and retraining; compute savings on data volume are the upside. Can fail because the guarantees hold only in a stylized asymptotic setting (pruning along a one-dimensional subspace, known noise model), and keeping only hardest examples is fragile under real label noise, where the hardest examples are often mislabeled.", "rationale": "The paper is a rigorous but purely theoretical analysis whose optimal strategy is proven only in a high-dimensional linearized regime, so the reader should validate hardest-example pruning on their own workload at small scale before adopting it in production.", "technique": "Hard-example data pruning via weighted ERM", "family": "treino_eficiente", "recommendation": "testar", "published": "2026-10-15", "as_of": "2026-10-18", "editorial_status": "indexed", "source_url": "https://arxiv.org/abs/2610.21201", "independent_implementations": 0, "total_implementations": 1, "stars": 4, "citations": 0, "claims": [ { "claim_id": "claim-01", "statement": "Replaces training on the full dataset with drastic pruning that retains only the hardest examples (as few as 2N) under weighted empirical risk minimization. Costs are a scoring/ranking pass over the training set to identify hard examples and retraining; compute savings on data volume are the upside. Can fail because the guarantees hold only in a stylized asymptotic setting (pruning along a one-dimensional subspace, known noise model), and keeping only hardest examples is fragile under real label noise, where the hardest examples are often mislabeled.", "basis": "inferred", "result": "", "baseline": "", "conditions": "", "source_url": "https://arxiv.org/abs/2610.21201", "source_page": null, "source_excerpt": "" }, { "claim_id": "claim-02", "statement": "Theoretical result: in the high-sample regime (P >> N), optimal pruning to the 2N hardest examples accelerates classification test-error decay from the classical sqrt(N/P) rate to the Bayes-optimal N/P rate. Asymptotic rate improvement, not a measured multiplicative factor.", "basis": "inferred", "result": "", "baseline": "", "conditions": "", "source_url": "https://arxiv.org/abs/2610.21201", "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": "" }