{ "schema_version": 1, "builder_plan": { "scenario": "An agent reads external documents or tool responses that may contain malicious instructions.", "change": "Replay attacks and legitimate lookalikes in an isolated copy of your agent workflow.", "baseline": "Your existing checks with the same tasks, permissions, and tool responses.", "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": "Successful attacks missed, legitimate actions blocked, detection delay, and review 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": "2609.37858", "title": "Storage Is Not Strategy: State-Conditioned Support Control for LLM Unlearning", "summary": "The method replaces fixed storage-localization-based parameter subset selection for localized LLM unlearning with an intervention-value ranking that is selectively re-evaluated during optimization when a calibrated probe triggers it. It costs additional probe evaluations and re-ranking computation during fine-tuning plus implementation complexity, though the paper shows plain LoRA already wins 35/36 targets, so the added machinery must beat a cheap LoRA baseline. Gains are benchmark-dependent (notably unstable for the GradDiff objective) and several reported margins are near zero, so adoption can yield no practical improvement on the reader's specific unlearning task.", "rationale": "LLM unlearning is a specialized privacy need and the evidence shows inconsistent, often marginal gains plus a strong cheap LoRA baseline, so it warrants a small controlled pilot only if unlearning is an actual production requirement.", "technique": "State-conditioned dynamic intervention re-ranking for LLM unlearning", "family": "seguranca_e_guardrails", "recommendation": "testar", "published": "2026-09-29", "as_of": "2026-10-18", "editorial_status": "indexed", "source_url": "https://arxiv.org/abs/2609.37858", "independent_implementations": 0, "total_implementations": 0, "stars": 0, "citations": null, "claims": [ { "claim_id": "claim-01", "statement": "The method replaces fixed storage-localization-based parameter subset selection for localized LLM unlearning with an intervention-value ranking that is selectively re-evaluated during optimization when a calibrated probe triggers it. It costs additional probe evaluations and re-ranking computation during fine-tuning plus implementation complexity, though the paper shows plain LoRA already wins 35/36 targets, so the added machinery must beat a cheap LoRA baseline. Gains are benchmark-dependent (notably unstable for the GradDiff objective) and several reported margins are near zero, so adoption can yield no practical improvement on the reader's specific unlearning task.", "basis": "inferred", "result": "", "baseline": "", "conditions": "", "source_url": "https://arxiv.org/abs/2609.37858", "source_page": null, "source_excerpt": "" }, { "claim_id": "claim-02", "statement": "On LACUNA, mean terminal utility improves over static selection in all six comparisons: NPO margins +0.431 to +0.848, SimNPO +0.503 to +0.571; on Natural-TOFU, positive margins in 19/20 comparisons, several near zero; GradDiff vs Static-IV shows six wins, six ties, mean paired gain +0.165.", "basis": "inferred", "result": "", "baseline": "", "conditions": "", "source_url": "https://arxiv.org/abs/2609.37858", "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": "" }