{
  "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": "2609.20152",
  "title": "MTVA-Bench: Evaluating the Language Model Inside Cascaded Voice Agents",
  "summary": "MTVA-Bench replaces end-to-end voice pipeline scoring and generic LLM benchmarks with an isolated evaluation of the language model under realistic call conditions (ASR errors, split utterances, script and language constraints), using an LLM caller, a mock backend, deterministic tool-call checks, and two citation-requiring LLM judges across 490 scenarios in 7 languages. Adopting it costs the engineering effort of integrating an agent harness and mock backends plus recurring LLM judge inference, with no model retraining required. It can fail through judge unreliability or citation gaming, limited scenario coverage relative to a real production domain, and overfitting to the benchmark's simulated caller rather than actual human callers.",
  "rationale": "For a small team running cascaded voice agents via third-party APIs, this benchmark offers a low-infrastructure way to select and regression-test models, but its simulated scenarios and LLM judges require validation against the team's real call distribution before trusted adoption.",
  "technique": "Benchmark for LLMs inside cascaded voice agents",
  "family": "avaliacao_de_agente",
  "recommendation": "testar",
  "published": "2026-09-17",
  "as_of": "2026-10-18",
  "editorial_status": "indexed",
  "source_url": "https://arxiv.org/abs/2609.20152",
  "independent_implementations": 3,
  "total_implementations": 4,
  "stars": 4,
  "citations": 0,
  "claims": [
    {
      "claim_id": "claim-01",
      "statement": "MTVA-Bench replaces end-to-end voice pipeline scoring and generic LLM benchmarks with an isolated evaluation of the language model under realistic call conditions (ASR errors, split utterances, script and language constraints), using an LLM caller, a mock backend, deterministic tool-call checks, and two citation-requiring LLM judges across 490 scenarios in 7 languages. Adopting it costs the engineering effort of integrating an agent harness and mock backends plus recurring LLM judge inference, with no model retraining required. It can fail through judge unreliability or citation gaming, limited scenario coverage relative to a real production domain, and overfitting to the benchmark's simulated caller rather than actual human callers.",
      "basis": "inferred",
      "result": "",
      "baseline": "",
      "conditions": "",
      "source_url": "https://arxiv.org/abs/2609.20152",
      "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": ""
}