{ "schema_version": 1, "builder_plan": { "scenario": "An agent looks good in demos, but you cannot tell which product tasks it fails.", "change": "Use the study's evaluation idea to inspect a small set of tasks with human-checked answers.", "baseline": "Your current evaluator and a manually reviewed reference set.", "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": "Agreement with human review, missed failures, false alarms, and evaluation cost per 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": "2610.22031", "title": "GraphACM: Graph-based Action Consistency Measurement for Uncertainty Quantification in LLM Agents", "summary": "Replaces single task-level consistency scores and verbalized confidence with a state-conditioned measure built by merging repeated rollouts into an action-labeled graph and computing node-level and global action consistency. Costs several full agent executions per task (multiplying API tokens and latency) plus implementation of state-equivalence merging, which is straightforward for textual state representations but nontrivial otherwise. Fails when the state-equivalence criterion mismatches the environment (fragmenting or over-merging the graph), when rollouts are too few for a stable estimate, or when the AUROC gains measured on ALFWorld, AppWorld, and WebShop do not transfer to the reader's domain.", "rationale": "It is a black-box, training-free reliability signal implementable with API access, but it requires multiple rollouts per task and benchmark-validated gains must be re-verified on the reader's own agent workload before production use.", "technique": "Graph-based action consistency for agent uncertainty estimation", "family": "avaliacao_de_agente", "recommendation": "testar", "published": "2026-10-15", "as_of": "2026-10-18", "editorial_status": "indexed", "source_url": "https://arxiv.org/abs/2610.22031", "independent_implementations": 1, "total_implementations": 2, "stars": 132, "citations": null, "claims": [ { "claim_id": "claim-01", "statement": "Replaces single task-level consistency scores and verbalized confidence with a state-conditioned measure built by merging repeated rollouts into an action-labeled graph and computing node-level and global action consistency. Costs several full agent executions per task (multiplying API tokens and latency) plus implementation of state-equivalence merging, which is straightforward for textual state representations but nontrivial otherwise. Fails when the state-equivalence criterion mismatches the environment (fragmenting or over-merging the graph), when rollouts are too few for a stable estimate, or when the AUROC gains measured on ALFWorld, AppWorld, and WebShop do not transfer to the reader's domain.", "basis": "inferred", "result": "", "baseline": "", "conditions": "", "source_url": "https://arxiv.org/abs/2610.22031", "source_page": null, "source_excerpt": "" }, { "claim_id": "claim-02", "statement": "Global consistency score achieves 0.811 average AUROC for task-success prediction, versus 0.729 for the strongest behavioral-consistency baseline and 0.769 for verbalized confidence, across three LLMs on ALFWorld, AppWorld, and WebShop.", "basis": "inferred", "result": "", "baseline": "", "conditions": "", "source_url": "https://arxiv.org/abs/2610.22031", "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": "" }