{ "schema_version": 1, "builder_plan": { "scenario": "An agent looks good in demos, but you cannot tell which product tasks it fails.", "change": "Use the study's evaluation idea to inspect a small set of tasks with human-checked answers.", "baseline": "Your current evaluator and a manually reviewed reference set.", "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": "Agreement with human review, missed failures, false alarms, and evaluation cost per task.", "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": "2609.25591", "title": "Evaluating Coding Agents on Kernel Exploit Generation", "summary": "KEX-bench replaces ad hoc bug-discovery evaluations of coding agents with VM-isolated, deterministically verified tasks that test whether agents can turn kernel crashes into exploit primitives. Adoption costs VM infrastructure per task, careful environment reproduction, and access to frontier or open-weight models under fixed tool-call budgets; it provides a measurement instrument, not a capability improvement. Results can fail to transfer because scores depend heavily on the availability of a reference PoC, on the specific agent-model pairing, and on a narrow 45-task CVE sample that may not represent a given production target.", "rationale": "This is a security evaluation benchmark rather than a deployable inference or training technique, so it matters only if the reader evaluates or procures coding agents for offensive-security tasks.", "technique": "KEX-bench: kernel exploit-generation agent benchmark", "family": "avaliacao_de_agente", "recommendation": "observar", "published": "2026-09-22", "as_of": "2026-10-18", "editorial_status": "indexed", "source_url": "https://arxiv.org/abs/2609.25591", "independent_implementations": 0, "total_implementations": 0, "stars": 0, "citations": null, "claims": [ { "claim_id": "claim-01", "statement": "KEX-bench replaces ad hoc bug-discovery evaluations of coding agents with VM-isolated, deterministically verified tasks that test whether agents can turn kernel crashes into exploit primitives. Adoption costs VM infrastructure per task, careful environment reproduction, and access to frontier or open-weight models under fixed tool-call budgets; it provides a measurement instrument, not a capability improvement. Results can fail to transfer because scores depend heavily on the availability of a reference PoC, on the specific agent-model pairing, and on a narrow 45-task CVE sample that may not represent a given production target.", "basis": "inferred", "result": "", "baseline": "", "conditions": "", "source_url": "https://arxiv.org/abs/2609.25591", "source_page": null, "source_excerpt": "" }, { "claim_id": "claim-02", "statement": "Strongest agent-model configuration solves 56.0% of Linux and 5.0% of Windows tasks without a reference PoC, rising to 68.9% of all 45 tasks with a reference PoC.", "basis": "inferred", "result": "", "baseline": "", "conditions": "", "source_url": "https://arxiv.org/abs/2609.25591", "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": "" }