{
  "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": "2610.29458",
  "title": "Harness Scaling for Generalizable Agentic Reinforcement Learning",
  "summary": "Harness Scaling replaces agentic RL performed through a single fixed harness with RL conducted across a factorized, systematically sampled set of harness configurations, which the authors report improves generalization to unseen white-box and black-box harnesses. Its cost is substantial: it requires RL training runs across multiple harness configurations, repeated for several model scales and algorithms, which demands training compute, harness engineering, and evaluation infrastructure well beyond a single 24 GB GPU or API-only setup. Adoption can fail if the factorized harness space omits the dimensions that matter for a deployment harness, if the balanced factorial subset is too small to induce transfer, or if the reader's underlying API model cannot be fine-tuned at all, in which case the method is inapplicable.",
  "rationale": "The technique is an RL training methodology requiring fine-tuning access and multi-configuration training runs that the reader's constrained infrastructure and API-only models cannot support, so it is relevant to monitor but not adoptable now.",
  "technique": "Harness Scaling: RL across diverse agent harnesses",
  "family": "uso_de_ferramenta",
  "recommendation": "observar",
  "published": "2026-10-21",
  "as_of": "2026-10-22",
  "editorial_status": "indexed",
  "source_url": "https://arxiv.org/abs/2610.29458",
  "independent_implementations": 4,
  "total_implementations": 5,
  "stars": 15,
  "citations": null,
  "claims": [
    {
      "claim_id": "claim-01",
      "statement": "Harness Scaling replaces agentic RL performed through a single fixed harness with RL conducted across a factorized, systematically sampled set of harness configurations, which the authors report improves generalization to unseen white-box and black-box harnesses. Its cost is substantial: it requires RL training runs across multiple harness configurations, repeated for several model scales and algorithms, which demands training compute, harness engineering, and evaluation infrastructure well beyond a single 24 GB GPU or API-only setup. Adoption can fail if the factorized harness space omits the dimensions that matter for a deployment harness, if the balanced factorial subset is too small to induce transfer, or if the reader's underlying API model cannot be fine-tuned at all, in which case the method is inapplicable.",
      "basis": "inferred",
      "result": "",
      "baseline": "",
      "conditions": "",
      "source_url": "https://arxiv.org/abs/2610.29458",
      "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": ""
}