{ "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": "2609.27336", "title": "CART: Closed-Loop Adaptive Red Teaming for Large Language Models", "summary": "CART replaces one-shot replay of a fixed red-team prompt set with a closed-loop Challenger-Target-Judge loop that adapts probes to emerging weaknesses while logging evidence and provenance. It costs Challenger and Judge inference per iteration (third-party APIs are viable), plus the engineering to separate roles, maintain probe diversity, and store auditable findings. It can fail because Challenger-Judge model choices bias which evidence surfaces, results describe discoverable failures rather than real-world failure rates, and unconstrained loops may drift into redundant or unrepresentative attacks without independent review.", "rationale": "Adaptive red teaming is implementable on API budgets and improves failure discovery over static replay, but the sensitivity to Challenger-Judge choices and lack of deployment-frequency grounding mean the reader should validate it on their own threat model before relying on it as a production gate.", "technique": "Closed-loop adaptive red teaming for LLMs", "family": "seguranca_e_guardrails", "recommendation": "testar", "published": "2026-09-23", "as_of": "2026-10-18", "editorial_status": "indexed", "source_url": "https://arxiv.org/abs/2609.27336", "independent_implementations": 0, "total_implementations": 1, "stars": 2, "citations": 0, "claims": [ { "claim_id": "claim-01", "statement": "CART replaces one-shot replay of a fixed red-team prompt set with a closed-loop Challenger-Target-Judge loop that adapts probes to emerging weaknesses while logging evidence and provenance. It costs Challenger and Judge inference per iteration (third-party APIs are viable), plus the engineering to separate roles, maintain probe diversity, and store auditable findings. It can fail because Challenger-Judge model choices bias which evidence surfaces, results describe discoverable failures rather than real-world failure rates, and unconstrained loops may drift into redundant or unrepresentative attacks without independent review.", "basis": "inferred", "result": "", "baseline": "", "conditions": "", "source_url": "https://arxiv.org/abs/2609.27336", "source_page": null, "source_excerpt": "" }, { "claim_id": "claim-02", "statement": "Discovers more failures and higher average risk than static seed replay for every Target with an available baseline, across three evaluation families (Frontier, JAH, Agentic); no multiplicative factor is reported.", "basis": "inferred", "result": "", "baseline": "", "conditions": "", "source_url": "https://arxiv.org/abs/2609.27336", "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": "" }