{
  "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": "2609.30012",
  "title": "Low-Cost Assays for Measuring Model Behavior Across Vendors and Releases",
  "summary": "Replaces bespoke, one-off evaluation harnesses and expensive human-coded transcript analysis with frozen public stimuli scored by exact match on clamped replies, LLM judges with per-code human-agreement reporting, or instrumented environments that log actions. Costs a few dollars per model per run plus the engineering effort to clamp outputs, maintain codebooks, and build instrumented sandboxes, with no GPU or training required. LLM-judge coding can silently drift from human judgment, clamped formats can distort natural behavior, and results are harness-dependent (the same model changed behavior across coding-agent harnesses), so comparisons confound model and scaffolding.",
  "rationale": "A small team on API budgets can adopt this immediately to cheaply and repeatedly track vendor and release behavior changes, which is exactly the constrained-infrastructure production setting.",
  "technique": "Frozen public stimulus batteries for cross-vendor behavioral evaluation",
  "family": "avaliacao_de_agente",
  "recommendation": "adotar",
  "published": "2026-09-24",
  "as_of": "2026-10-18",
  "editorial_status": "indexed",
  "source_url": "https://arxiv.org/abs/2609.30012",
  "independent_implementations": 0,
  "total_implementations": 0,
  "stars": 0,
  "citations": 0,
  "claims": [
    {
      "claim_id": "claim-01",
      "statement": "Replaces bespoke, one-off evaluation harnesses and expensive human-coded transcript analysis with frozen public stimuli scored by exact match on clamped replies, LLM judges with per-code human-agreement reporting, or instrumented environments that log actions. Costs a few dollars per model per run plus the engineering effort to clamp outputs, maintain codebooks, and build instrumented sandboxes, with no GPU or training required. LLM-judge coding can silently drift from human judgment, clamped formats can distort natural behavior, and results are harness-dependent (the same model changed behavior across coding-agent harnesses), so comparisons confound model and scaffolding.",
      "basis": "inferred",
      "result": "",
      "baseline": "",
      "conditions": "",
      "source_url": "https://arxiv.org/abs/2609.30012",
      "source_page": null,
      "source_excerpt": ""
    },
    {
      "claim_id": "claim-02",
      "statement": "Each frozen assay runs identically on a cross-vendor panel at a few dollars per model or less, including LLM-judge coding and instrumented agent environments; no multiplicative improvement factor over an alternative is reported.",
      "basis": "inferred",
      "result": "",
      "baseline": "",
      "conditions": "",
      "source_url": "https://arxiv.org/abs/2609.30012",
      "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": ""
}