{
  "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": "2609.24174",
  "title": "CREDO: Variance-Guided Rubric Evolution for Replay-Corrected Credit Assignment",
  "summary": "CREDO replaces naive terminal-reward credit assignment in RL training of long-horizon language agents with a learned credit head over rubric features, corrected by selectively replayed counterfactual continuations. It requires a resettable training environment, a frozen judge model, an extra credit head, and an expected replay budget, all of which add training complexity and compute on top of PPO. It can fail when rubrics are misspecified, the credit head is biased under distribution shift, or replay coverage is sparse, and the authors explicitly make no empirical superiority claim on language-agent benchmarks.",
  "rationale": "This is a preliminary RL training method requiring resettable environments, counterfactual rollouts, and PPO training infrastructure that exceeds the reader's constrained setup, with no demonstrated empirical gains to justify adoption.",
  "technique": "Variance-guided rubric credit assignment with selective replays",
  "family": "treino_eficiente",
  "recommendation": "observar",
  "published": "2026-09-21",
  "as_of": "2026-10-18",
  "editorial_status": "indexed",
  "source_url": "https://arxiv.org/abs/2609.24174",
  "independent_implementations": 0,
  "total_implementations": 0,
  "stars": 0,
  "citations": 0,
  "claims": [
    {
      "claim_id": "claim-01",
      "statement": "CREDO replaces naive terminal-reward credit assignment in RL training of long-horizon language agents with a learned credit head over rubric features, corrected by selectively replayed counterfactual continuations. It requires a resettable training environment, a frozen judge model, an extra credit head, and an expected replay budget, all of which add training complexity and compute on top of PPO. It can fail when rubrics are misspecified, the credit head is biased under distribution shift, or replay coverage is sparse, and the authors explicitly make no empirical superiority claim on language-agent benchmarks.",
      "basis": "inferred",
      "result": "",
      "baseline": "",
      "conditions": "",
      "source_url": "https://arxiv.org/abs/2609.24174",
      "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": ""
}