{ "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": "2610.20939", "title": "MAI-Gym: An Open All-in-One Platform for GUI Agent Evaluation and RL Training across Mobile and Computer Environments", "summary": "It replaces ad-hoc GUI-agent evaluation and per-benchmark training pipelines with one shared rollout system spanning mobile, computer, and web, plus automated task and mocked-app synthesis. The cost is infrastructure: results depend on thousands of parallel environment instances and RL training of a 9B model, neither of which fits a single 24 GB GPU or a limited cloud budget. Failure modes include reward hacking on execution-validated verifiers, imperfect mock-to-real transfer (48.1% success on the physical phone leaves most tasks failing), and evaluator-environment drift if the synthesis pipeline does not match real target apps.", "rationale": "The platform's gains come from scale and RL training that constrained infrastructure cannot reproduce, though its released benchmarks and verified tasks are worth monitoring for evaluation use.", "technique": "Scalable online RL platform for GUI agents", "family": "avaliacao_de_agente", "recommendation": "observar", "published": "2026-10-14", "as_of": "2026-10-18", "editorial_status": "indexed", "source_url": "https://arxiv.org/abs/2610.20939", "independent_implementations": 2, "total_implementations": 4, "stars": 158, "citations": 0, "claims": [ { "claim_id": "claim-01", "statement": "It replaces ad-hoc GUI-agent evaluation and per-benchmark training pipelines with one shared rollout system spanning mobile, computer, and web, plus automated task and mocked-app synthesis. The cost is infrastructure: results depend on thousands of parallel environment instances and RL training of a 9B model, neither of which fits a single 24 GB GPU or a limited cloud budget. Failure modes include reward hacking on execution-validated verifiers, imperfect mock-to-real transfer (48.1% success on the physical phone leaves most tasks failing), and evaluator-environment drift if the synthesis pipeline does not match real target apps.", "basis": "inferred", "result": "", "baseline": "", "conditions": "", "source_url": "https://arxiv.org/abs/2610.20939", "source_page": null, "source_excerpt": "" }, { "claim_id": "claim-02", "statement": "RL-trained MAI-Gym-9B reaches 78.4% on MobileWorld and 76.2% on OSWorld-Verified, best among open GUI agents under 10B; mocked-app RL transfers to a real phone, raising success from 28.8% to 48.1%.", "basis": "inferred", "result": "", "baseline": "", "conditions": "", "source_url": "https://arxiv.org/abs/2610.20939", "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": "" }