{ "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": "2610.26469", "title": "VeriCUA: Learning Proactive Verification for Reliable GUI Agent Rewards", "summary": "It replaces prompted foundation-model judges (4B to GPT-5.4 scale, plateaued at 68%-70% agreement) with a small trainable Judge, as small as 1B, that learns what to check and when to stop via online RL while a frozen Probe executes checks in the live environment. The cost is an online RL training loop requiring interactive OSWorld-style environments, single-attempt verification per terminal state, and return normalization across states, plus engineering to couple the Judge to a live GUI sandbox. It can fail when target environments differ from the training distribution despite the reported cross-benchmark transfer, when reward agreement gains do not translate to Actor task success on a specific workload, or when the team cannot maintain the live-environment training infrastructure.", "rationale": "A 1B trainable verifier fitting on one 24 GB GPU is within reach, but the online RL training against live GUI environments and the need to validate agreement gains on the reader's own agent tasks make workload-specific testing necessary before production adoption.", "technique": "RL-trained proactive verification judge for GUI agent rewards", "family": "avaliacao_de_agente", "recommendation": "testar", "published": "2026-10-19", "as_of": "2026-10-20", "editorial_status": "indexed", "source_url": "https://arxiv.org/abs/2610.26469", "independent_implementations": 2, "total_implementations": 3, "stars": 132, "citations": null, "claims": [ { "claim_id": "claim-01", "statement": "It replaces prompted foundation-model judges (4B to GPT-5.4 scale, plateaued at 68%-70% agreement) with a small trainable Judge, as small as 1B, that learns what to check and when to stop via online RL while a frozen Probe executes checks in the live environment. The cost is an online RL training loop requiring interactive OSWorld-style environments, single-attempt verification per terminal state, and return normalization across states, plus engineering to couple the Judge to a live GUI sandbox. It can fail when target environments differ from the training distribution despite the reported cross-benchmark transfer, when reward agreement gains do not translate to Actor task success on a specific workload, or when the team cannot maintain the live-environment training infrastructure.", "basis": "inferred", "result": "", "baseline": "", "conditions": "", "source_url": "https://arxiv.org/abs/2610.26469", "source_page": null, "source_excerpt": "" }, { "claim_id": "claim-02", "statement": "Surpasses few-shot GPT-5.4 by 5.36 points macro-average reward agreement with 47.6%-59.8% fewer Probe tokens; as an Actor RL reward it raises task success by 3.21 points overall and 5.40 on held-out multi-application tasks.", "basis": "inferred", "result": "", "baseline": "", "conditions": "", "source_url": "https://arxiv.org/abs/2610.26469", "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": "" }