{
  "schema_version": 1,
  "builder_plan": {
    "scenario": "An assistant chooses the wrong API or sends arguments your product cannot use.",
    "change": "Apply the proposed tool-selection or argument-generation change in a sandbox.",
    "baseline": "Your current agent with the same tool schemas and task budget.",
    "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": "Correct completed actions, invalid calls, retries, latency, and cost per completed 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": "2609.24985",
  "title": "Critical-State RL: Diagnosing Trainable States for Multi-Turn Tool Use",
  "summary": "The method replaces reward-variation-based or uniform selection of which model calls to train in multi-turn tool-use trajectories with a diagnostic that filters states by causal reward attribution and improvement headroom, then trains only those states as contextual bandits. It costs nested sampling rollouts per candidate state, task-defined candidate calls and local rewards, and a reference policy, adding pipeline complexity though the resulting training itself is lightweight. It can fail when local rewards are unavailable or misspecified, when the diagnostic selects states that do not transfer to the deployed model or task distribution, or when continuation-noise estimates are unreliable at small sample sizes.",
  "rationale": "The recipe targets a concrete small-team problem, improving multi-turn tool use without training full trajectories, but its gains depend on task-specific local rewards and candidate-state definitions that require workload-specific validation before adoption.",
  "technique": "Critical-State RL for multi-turn tool use",
  "family": "uso_de_ferramenta",
  "recommendation": "testar",
  "published": "2026-09-21",
  "as_of": "2026-10-18",
  "editorial_status": "indexed",
  "source_url": "https://arxiv.org/abs/2609.24985",
  "independent_implementations": 0,
  "total_implementations": 0,
  "stars": 0,
  "citations": 0,
  "claims": [
    {
      "claim_id": "claim-01",
      "statement": "The method replaces reward-variation-based or uniform selection of which model calls to train in multi-turn tool-use trajectories with a diagnostic that filters states by causal reward attribution and improvement headroom, then trains only those states as contextual bandits. It costs nested sampling rollouts per candidate state, task-defined candidate calls and local rewards, and a reference policy, adding pipeline complexity though the resulting training itself is lightweight. It can fail when local rewards are unavailable or misspecified, when the diagnostic selects states that do not transfer to the deployed model or task distribution, or when continuation-noise estimates are unreliable at small sample sizes.",
      "basis": "inferred",
      "result": "",
      "baseline": "",
      "conditions": "",
      "source_url": "https://arxiv.org/abs/2609.24985",
      "source_page": null,
      "source_excerpt": ""
    },
    {
      "claim_id": "claim-02",
      "statement": "Training diagnostic-selected states improves BFCL v4 performance by about 14 percentage points on the missing-function task, while training alternative states leaves performance flat or worse.",
      "basis": "inferred",
      "result": "",
      "baseline": "",
      "conditions": "",
      "source_url": "https://arxiv.org/abs/2609.24985",
      "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": ""
}