{ "schema_version": 1, "builder_plan": { "scenario": "A workflow needs specialist review, but extra agents may add more cost than useful work.", "change": "Test the proposed division of work against a single-agent implementation.", "baseline": "One agent with the same tools and an equal total token and time budget.", "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": "Accepted outputs, coordination failures, wall-clock time, and total cost across all agents.", "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": "2610.13494", "title": "Humanize MLE Agents: Multi-Agent Session Alternation Beyond Single-Agent Limits", "summary": "HMA replaces single long-horizon agent sessions with custom context-engineering harnesses by alternating two agents over a shared workspace, capping each session at k experiments and resetting context between sessions. The cost is orchestration logic and duplicated token spend across sessions (each agent re-reads workspace state), though the paper reports a 4x reduction in wall-clock budget relative to the baseline; it requires API access to two capable LLM agents, which fits constrained infrastructure. It can fail when the two agents lack complementary strengths, when workspace state summaries lose critical experiment context across resets, or when the task is short enough that single-session deceleration never materializes.", "rationale": "The method is implementable with third-party APIs and no custom training, and shows gains across heterogeneous agent pairs, but the reported +6.4 point average improvement is modest relative to per-task variance and the theoretical guarantees are asymptotic, so adoption should be validated on the reader's own MLE workloads first.", "technique": "Session-alternating multi-agent MLE workflow", "family": "orquestracao_multiagente", "recommendation": "testar", "published": "2026-10-09", "as_of": "2026-10-18", "editorial_status": "indexed", "source_url": "https://arxiv.org/abs/2610.13494", "independent_implementations": 1, "total_implementations": 2, "stars": 132, "citations": null, "claims": [ { "claim_id": "claim-01", "statement": "HMA replaces single long-horizon agent sessions with custom context-engineering harnesses by alternating two agents over a shared workspace, capping each session at k experiments and resetting context between sessions. The cost is orchestration logic and duplicated token spend across sessions (each agent re-reads workspace state), though the paper reports a 4x reduction in wall-clock budget relative to the baseline; it requires API access to two capable LLM agents, which fits constrained infrastructure. It can fail when the two agents lack complementary strengths, when workspace state summaries lose critical experiment context across resets, or when the task is short enough that single-session deceleration never materializes.", "basis": "inferred", "result": "", "baseline": "", "conditions": "", "source_url": "https://arxiv.org/abs/2610.13494", "source_page": null, "source_excerpt": "" }, { "claim_id": "claim-02", "statement": "HMA reaches a 78.2% any-medal rate on 75 MLE-bench Kaggle tasks, 8.0 points above the strongest open-source baseline, using a quarter of its time budget; across six agent-pair configurations it improves any-medal rate by 6.4 points on average over constituent baselines.", "basis": "inferred", "result": "", "baseline": "", "conditions": "", "source_url": "https://arxiv.org/abs/2610.13494", "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": "" }