{
  "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": "2610.21109",
  "title": "Kolab: Benchmarking Human-Agent Collaboration in Long-Horizon Workflows",
  "summary": "Kolab replaces automation-centric agent benchmarks (single-attempt task completion) with 115+ hours of real professional workflow recordings that test two new capabilities: forecasting whether experts adopt, modify, or reject agent contributions, and assessing collaboration quality without constant feedback. Adopting it costs integration effort for a benchmark that evaluates collaboration-judgment capabilities most production systems do not implement, and its findings show frontier models overestimate collaboration quality regardless of context length, so scores offer limited actionable tuning signal. It can fail as an adoption guide because it measures a meta-capability (judging collaboration) rather than the task-level reliability a small team with one GPU actually needs to validate.",
  "rationale": "This is a diagnostic benchmark for a capability (modeling human collaboration) that is not yet deployable or improvable on constrained infrastructure, so tracking its findings is more appropriate than integrating it into a production evaluation pipeline.",
  "technique": "Human-agent collaboration benchmark from professional screen recordings",
  "family": "avaliacao_de_agente",
  "recommendation": "observar",
  "published": "2026-10-15",
  "as_of": "2026-10-18",
  "editorial_status": "indexed",
  "source_url": "https://arxiv.org/abs/2610.21109",
  "independent_implementations": 1,
  "total_implementations": 2,
  "stars": 14,
  "citations": 0,
  "claims": [
    {
      "claim_id": "claim-01",
      "statement": "Kolab replaces automation-centric agent benchmarks (single-attempt task completion) with 115+ hours of real professional workflow recordings that test two new capabilities: forecasting whether experts adopt, modify, or reject agent contributions, and assessing collaboration quality without constant feedback. Adopting it costs integration effort for a benchmark that evaluates collaboration-judgment capabilities most production systems do not implement, and its findings show frontier models overestimate collaboration quality regardless of context length, so scores offer limited actionable tuning signal. It can fail as an adoption guide because it measures a meta-capability (judging collaboration) rather than the task-level reliability a small team with one GPU actually needs to validate.",
      "basis": "inferred",
      "result": "",
      "baseline": "",
      "conditions": "",
      "source_url": "https://arxiv.org/abs/2610.21109",
      "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": ""
}