{ "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": "2609.40063", "title": "LARC: Low-Rank Adaptive Residual Connections for Learning in Frozen Models", "summary": "LARC replaces full fine-tuning or prompt-based adaptation with a 12,288-parameter rank-4 residual (h+BAh) on a frozen 1B model, using a slow learned initialization and a fast state that updates from feedback and resets. The compute and memory cost is negligible, but it requires a feedback-gradient signal at inference time, reset-policy management, and a separate selection rule. It can fail through online-update degradation: retaining updates worsened downstream Brier loss, updates showed same-batch non-descent, and benefits were inconsistent across only three paired seeds.", "rationale": "The evidence is a small-scale development study with mixed results, including a demonstrated case where the core online mechanism degraded performance, so it warrants monitoring rather than adoption or small-scale testing for production use.", "technique": "Low-rank residual adapter with resettable fast state", "family": "treino_eficiente", "recommendation": "observar", "published": "2026-09-30", "as_of": "2026-10-18", "editorial_status": "indexed", "source_url": "https://arxiv.org/abs/2609.40063", "independent_implementations": 0, "total_implementations": 0, "stars": 0, "citations": null, "claims": [ { "claim_id": "claim-01", "statement": "LARC replaces full fine-tuning or prompt-based adaptation with a 12,288-parameter rank-4 residual (h+BAh) on a frozen 1B model, using a slow learned initialization and a fast state that updates from feedback and resets. The compute and memory cost is negligible, but it requires a feedback-gradient signal at inference time, reset-policy management, and a separate selection rule. It can fail through online-update degradation: retaining updates worsened downstream Brier loss, updates showed same-batch non-descent, and benefits were inconsistent across only three paired seeds.", "basis": "inferred", "result": "", "baseline": "", "conditions": "", "source_url": "https://arxiv.org/abs/2609.40063", "source_page": null, "source_excerpt": "" }, { "claim_id": "claim-02", "statement": "Two feedback-gradient steps on a rank-4 adapter reduced expected query execution error by 24.65 and 36.65 percentage points relative to static and post-adaptation initializations, but retaining online updates in chronological replay raised half-Brier loss from 0.1274 to 0.1808.", "basis": "inferred", "result": "", "baseline": "", "conditions": "", "source_url": "https://arxiv.org/abs/2609.40063", "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": "" }