{ "schema_version": 1, "builder_plan": { "scenario": "You already need task-specific training and want to reduce its memory or compute cost.", "change": "Change only the training technique on a small controlled run after checking prerequisites.", "baseline": "Your current training recipe with the same model, data split, and evaluation.", "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": "Held-out quality, peak memory, training time, and full compute cost to the target quality.", "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": "2609.39325", "title": "WorkGenesis: Building the Worlds That Teach Agents to Work", "summary": "WorkGenesis replaces expert-authored occupational training tasks with an automated pipeline that retrieves real public files guided by O*NET knowledge, synthesizes the surrounding task context and itemwise rubric, and iteratively repairs defects via execution-guided consistency checks. The cost is substantial API inference for retrieval, synthesis, rendering reference deliverables, and multi-round auditing, plus engineering effort to build the pipeline; the reader cannot replicate the 35B SFT run on a 24 GB GPU and would need API fine-tuning or a smaller model. Failures include residual rubric or task defects passing the automated audit, distribution mismatch between synthesized tasks and the reader's actual workload, and benchmark gains that may not transfer to domain-specific production tasks.", "rationale": "The data-generation pipeline is reproducible with third-party APIs at modest cost, but its value for the reader's specific agent workload is unproven and the reported training setup exceeds local hardware, so a small-scale pilot is warranted before adoption.", "technique": "Evidence-grounded synthetic work-task generation with execution-based verification", "family": "treino_eficiente", "recommendation": "testar", "published": "2026-09-30", "as_of": "2026-10-18", "editorial_status": "indexed", "source_url": "https://arxiv.org/abs/2609.39325", "independent_implementations": 0, "total_implementations": 0, "stars": 0, "citations": 0, "claims": [ { "claim_id": "claim-01", "statement": "WorkGenesis replaces expert-authored occupational training tasks with an automated pipeline that retrieves real public files guided by O*NET knowledge, synthesizes the surrounding task context and itemwise rubric, and iteratively repairs defects via execution-guided consistency checks. The cost is substantial API inference for retrieval, synthesis, rendering reference deliverables, and multi-round auditing, plus engineering effort to build the pipeline; the reader cannot replicate the 35B SFT run on a 24 GB GPU and would need API fine-tuning or a smaller model. Failures include residual rubric or task defects passing the automated audit, distribution mismatch between synthesized tasks and the reader's actual workload, and benchmark gains that may not transfer to domain-specific production tasks.", "basis": "inferred", "result": "", "baseline": "", "conditions": "", "source_url": "https://arxiv.org/abs/2609.39325", "source_page": null, "source_excerpt": "" }, { "claim_id": "claim-02", "statement": "A 35B model SFT-trained on 20K WorkGenesis-synthesized tasks scores 31.00 versus 24.79 average for comparable baselines across GDPvalAA-v2, APEX-Agents-AA, and JobBench, and reportedly surpasses DeepSeek-V4-Pro-Preview on these benchmarks.", "basis": "inferred", "result": "", "baseline": "", "conditions": "", "source_url": "https://arxiv.org/abs/2609.39325", "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": "" }