{
  "schema_version": 1,
  "builder_plan": {
    "scenario": "A workflow needs specialist review, but extra agents may add more cost than useful work.",
    "change": "Test the proposed division of work against a single-agent implementation.",
    "baseline": "One agent with the same tools and an equal total token and time 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": "Accepted outputs, coordination failures, wall-clock time, and total cost across all agents.",
    "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": "2610.28361",
  "title": "FALCA: Falsification-Aware Reinforcement Learning with Hierarchical Credit Assignment for Long-Horizon Deep Search Agents",
  "summary": "Replaces single-agent iterative deep search with a propose-falsify loop in which independent Falsifier agents seek counterevidence and revise the research state, optionally reinforced by SFT warmup and RL with hierarchical credit assignment. The inference-time cost is roughly doubled LLM calls per step plus orchestration overhead, and the training component requires parallel rollouts and staged reward assignment that are expensive on a 24 GB GPU and likely only feasible via third-party APIs. The loop can fail if Falsifiers share the backbone's blind spots and rubber-stamp candidates, if falsification reports are noisy enough to destabilize the research state, or if hierarchical credit assignment misattributes rewards across stages.",
  "rationale": "The harness-level Searcher-Falsifier pattern is cheap to prototype against API models and may reduce error accumulation, but the claimed gains are unquantified and the RL component is impractical on the reader's budget, so adoption should be gated on workload-specific evaluation.",
  "technique": "Searcher-Falsifier architecture with falsification-aware RL",
  "family": "orquestracao_multiagente",
  "recommendation": "testar",
  "published": "2026-10-20",
  "as_of": "2026-10-21",
  "editorial_status": "indexed",
  "source_url": "https://arxiv.org/abs/2610.28361",
  "independent_implementations": 3,
  "total_implementations": 4,
  "stars": 134,
  "citations": null,
  "claims": [
    {
      "claim_id": "claim-01",
      "statement": "Replaces single-agent iterative deep search with a propose-falsify loop in which independent Falsifier agents seek counterevidence and revise the research state, optionally reinforced by SFT warmup and RL with hierarchical credit assignment. The inference-time cost is roughly doubled LLM calls per step plus orchestration overhead, and the training component requires parallel rollouts and staged reward assignment that are expensive on a 24 GB GPU and likely only feasible via third-party APIs. The loop can fail if Falsifiers share the backbone's blind spots and rubber-stamp candidates, if falsification reports are noisy enough to destabilize the research state, or if hierarchical credit assignment misattributes rewards across stages.",
      "basis": "inferred",
      "result": "",
      "baseline": "",
      "conditions": "",
      "source_url": "https://arxiv.org/abs/2610.28361",
      "source_page": null,
      "source_excerpt": ""
    },
    {
      "claim_id": "claim-02",
      "statement": "Improves long-horizon deep search performance over backbone and harness ablations on five search benchmarks, with reduced confirmation bias and error accumulation; no quantified figure is given in the abstract.",
      "basis": "inferred",
      "result": "",
      "baseline": "",
      "conditions": "",
      "source_url": "https://arxiv.org/abs/2610.28361",
      "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": ""
}