{ "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": "2609.25237", "title": "Trains but Doesn't Learn: A Post-Training Delivery Benchmark for LLM Agents as Forward-Deployed Engineers", "summary": "The method replaces ad hoc post-training validation and metric-only agent benchmarks with a governed ten-stage delivery plane where an oracle scores each stage from platform-recorded facts, plus an acceptance gate that rejects runs whose delivered model does not beat the base. It costs an extra evaluation stage per run, calibration of the corruption detector on known-bad runs, and the operational complexity of recording platform facts for every pipeline stage. It can fail if the gate's baselines or detector thresholds do not transfer to the team's hardware and task mix, and because the benchmark's agent rankings depend on proprietary frontier models that will change, its comparative conclusions may age quickly.", "rationale": "The paper contributes an evaluation and governance framework rather than a performance technique, but its acceptance-gate and corruption-detection patterns are directly applicable to small-team fine-tuning pipelines and warrant workload-specific validation.", "technique": "Governed delivery benchmark for post-training LLM agents", "family": "avaliacao_de_agente", "recommendation": "testar", "published": "2026-09-21", "as_of": "2026-10-18", "editorial_status": "indexed", "source_url": "https://arxiv.org/abs/2609.25237", "independent_implementations": 0, "total_implementations": 0, "stars": 0, "citations": 0, "claims": [ { "claim_id": "claim-01", "statement": "The method replaces ad hoc post-training validation and metric-only agent benchmarks with a governed ten-stage delivery plane where an oracle scores each stage from platform-recorded facts, plus an acceptance gate that rejects runs whose delivered model does not beat the base. It costs an extra evaluation stage per run, calibration of the corruption detector on known-bad runs, and the operational complexity of recording platform facts for every pipeline stage. It can fail if the gate's baselines or detector thresholds do not transfer to the team's hardware and task mix, and because the benchmark's agent rankings depend on proprietary frontier models that will change, its comparative conclusions may age quickly.", "basis": "inferred", "result": "", "baseline": "", "conditions": "", "source_url": "https://arxiv.org/abs/2609.25237", "source_page": null, "source_excerpt": "" }, { "claim_id": "claim-02", "statement": "An operator-run acceptance gate catches every 'trains but doesn't learn' run before payment, and a detector calibrated on known-corrupted runs flags severe corruption mid-run; no improvement factor is reported.", "basis": "inferred", "result": "", "baseline": "", "conditions": "", "source_url": "https://arxiv.org/abs/2609.25237", "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": "" }