{
  "schema_version": 1,
  "builder_plan": {
    "scenario": "You already need task-specific training and want to reduce its memory or compute cost.",
    "change": "Change only the training technique on a small controlled run after checking prerequisites.",
    "baseline": "Your current training recipe with the same model, data split, and evaluation.",
    "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": "Held-out quality, peak memory, training time, and full compute cost to the target quality.",
    "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": "2610.13438",
  "title": "Rollout-Time Intervention: Unlocking Agentic Reinforcement Learning Beyond Policy Limits",
  "summary": "RTI replaces purely on-policy exploration in agentic RL by injecting reward-rubric evidence during rollout to convert failed trajectories into successful ones, then filtering interventions with excessive distributional shift before policy optimization. It costs additional rollout compute for intervention and filtering, requires well-specified evaluation rubrics per task, and adds off-policy correction logic to the training pipeline. It can fail when rubrics are unavailable or noisy, when filtering admits trajectories whose distribution shift destabilizes training, or when the resulting training still demands more GPU memory than a single 24 GB card provides for the chosen model size.",
  "rationale": "The method targets a real failure mode of small-scale agentic RL (sparse success on hard tasks) and is implementable with modest models and rubric-based rewards, but its benefit depends on task-specific rubric quality and off-policy filtering that must be validated on the reader's own workload.",
  "technique": "Rollout-time intervention for agentic RL",
  "family": "treino_eficiente",
  "recommendation": "testar",
  "published": "2026-10-09",
  "as_of": "2026-10-18",
  "editorial_status": "indexed",
  "source_url": "https://arxiv.org/abs/2610.13438",
  "independent_implementations": 2,
  "total_implementations": 3,
  "stars": 449,
  "citations": null,
  "claims": [
    {
      "claim_id": "claim-01",
      "statement": "RTI replaces purely on-policy exploration in agentic RL by injecting reward-rubric evidence during rollout to convert failed trajectories into successful ones, then filtering interventions with excessive distributional shift before policy optimization. It costs additional rollout compute for intervention and filtering, requires well-specified evaluation rubrics per task, and adds off-policy correction logic to the training pipeline. It can fail when rubrics are unavailable or noisy, when filtering admits trajectories whose distribution shift destabilizes training, or when the resulting training still demands more GPU memory than a single 24 GB card provides for the chosen model size.",
      "basis": "inferred",
      "result": "",
      "baseline": "",
      "conditions": "",
      "source_url": "https://arxiv.org/abs/2610.13438",
      "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": ""
}