{
  "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": "2610.29822",
  "title": "UserSWE: Evaluating Coding Agents in Multi-Turn Conversations with Dynamic Users",
  "summary": "UserSWE replaces single-turn issue-report benchmarks with multi-turn coding tasks in which a persona-driven simulated user reveals information progressively and execution-grounded tests gate verdicts after every turn. Adopting it costs the construction or reuse of its task pipeline plus API or compute spend for running agents, the simulated user, and the dual-family verdict agents across up to eight turns per task. It can fail to transfer to a specific production setting because the 72 tasks and 42 repositories may not match the reader's codebase, and simulated-user behavior calibrated on past sessions may diverge from real user interactions.",
  "rationale": "The reader selecting or comparing coding agents for production can use this benchmark to avoid mis-ranking agents whose single-turn scores do not predict multi-turn interactive performance, but its small task set requires validation against the reader's actual workload before driving adoption decisions.",
  "technique": "Multi-turn benchmark for coding agents with simulated users",
  "family": "avaliacao_de_agente",
  "recommendation": "testar",
  "published": "2026-10-21",
  "as_of": "2026-10-25",
  "editorial_status": "indexed",
  "source_url": "https://arxiv.org/abs/2610.29822",
  "independent_implementations": 0,
  "total_implementations": 1,
  "stars": 0,
  "citations": 0,
  "claims": [
    {
      "claim_id": "claim-01",
      "statement": "UserSWE replaces single-turn issue-report benchmarks with multi-turn coding tasks in which a persona-driven simulated user reveals information progressively and execution-grounded tests gate verdicts after every turn. Adopting it costs the construction or reuse of its task pipeline plus API or compute spend for running agents, the simulated user, and the dual-family verdict agents across up to eight turns per task. It can fail to transfer to a specific production setting because the 72 tasks and 42 repositories may not match the reader's codebase, and simulated-user behavior calibrated on past sessions may diverge from real user interactions.",
      "basis": "inferred",
      "result": "",
      "baseline": "",
      "conditions": "",
      "source_url": "https://arxiv.org/abs/2610.29822",
      "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": ""
}