{ "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": "2609.15938", "title": "HypoEvolve: Genetic Algorithms Enable Multi-Agent LLMs to Discover Scientific Hypotheses", "summary": "HypoEvolve replaces single-pass or loosely coordinated multi-agent hypothesis generation with a generational genetic algorithm in which specialized LLM agents critique, revise, and select a population of scientific hypotheses. It costs substantial API-call volume per generation (population size times generations times agent roles), added orchestration complexity, and requires domain-specific external evaluators such as DepMap and Open Targets. It can fail when no validated external scoring measure exists for the target domain, when the fitness signal rewards plausible-sounding but incorrect hypotheses, or when API costs grow faster than the quality gain justifies.", "rationale": "The method runs entirely through third-party LLM APIs and improves output quality on a verifiable metric, but its value depends on having a trustworthy external evaluator, so the reader should validate it on their own workload before production adoption.", "technique": "Genetic-algorithm orchestration of LLM hypothesis agents", "family": "orquestracao_multiagente", "recommendation": "testar", "published": "2026-09-14", "as_of": "2026-10-18", "editorial_status": "indexed", "source_url": "https://arxiv.org/abs/2609.15938", "independent_implementations": 0, "total_implementations": 0, "stars": 0, "citations": 0, "claims": [ { "claim_id": "claim-01", "statement": "HypoEvolve replaces single-pass or loosely coordinated multi-agent hypothesis generation with a generational genetic algorithm in which specialized LLM agents critique, revise, and select a population of scientific hypotheses. It costs substantial API-call volume per generation (population size times generations times agent roles), added orchestration complexity, and requires domain-specific external evaluators such as DepMap and Open Targets. It can fail when no validated external scoring measure exists for the target domain, when the fitness signal rewards plausible-sounding but incorrect hypotheses, or when API costs grow faster than the quality gain justifies.", "basis": "inferred", "result": "", "baseline": "", "conditions": "", "source_url": "https://arxiv.org/abs/2609.15938", "source_page": null, "source_excerpt": "" }, { "claim_id": "claim-02", "statement": "DepMap selectivity reaches 0.171 versus 0.115 for the strongest of six baselines across 34 cancer types, with gains over single-pass generation generalizing to held-out cancer types.", "basis": "inferred", "result": "", "baseline": "", "conditions": "", "source_url": "https://arxiv.org/abs/2609.15938", "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": [ { "anchor": "S3.E2", "label": "(2)", "section": "3 Method", "role": "proposed_method", "latex": "Q_{t}=P_{t-1}\\uplus O_{t},\\qquad f_{t}=\\operatorname{Score}(Q_{t}),\\qquad P_{t}=\\operatorname{Top}_{\\mu}(Q_{t};f_{t}).", "mathml": "Qt=Pt−1⊎Ot,ft=Score⁡(Qt),Pt=Topμ⁡(Qt;ft).", "context": "Algorithm Overview. HypoEvolve separates reasoning over scientific content from the population update that coordinates it. Agents supply hypothesis generation, semantic variation, and comparative fitness; the genetic algorithm specifies how these outputs change the population [18, 8]. A generation agent initializes P0 from retrieved literature. At generation t, selected parents produce λ offspring Ot through LLM-based crossover and mutation. A pairwise scorer evaluates parents and offspring together, and a deterministic supervisor retains the top μ candidates,", "headline": "", "explanation": "", "objective": "", "symbols": {}, "evidence": "", "evidence_url": "", "source_url": "" }, { "anchor": "S3.E1", "label": "(1)", "section": "3 Method", "role": "proposed_method", "latex": "h^{*}=\\arg\\max_{h\\in P_{G}}f_{G}(h).", "mathml": "h∗=arg⁡maxh∈PG​fG​(h).", "context": "Problem Formulation. Given a natural-language research goal g, we seek hypotheses that address the goal with scientifically grounded explanations and potentially new insights. Each candidate h is a structured document containing a title, summary, hypothesis statement, and supporting rationale. We formulate discovery as a finite population search with μ retained candidates, λ offspring per generation, and a horizon of G generations. Let Pt denote the population at generation t and ft the fitness inferred from task-specific comparisons in that generation. The search returns the highest-fitness hypothesis in the final population,", "headline": "", "explanation": "", "objective": "", "symbols": {}, "evidence": "", "evidence_url": "", "source_url": "" }, { "anchor": "S3.E3", "label": "(3)", "section": "3 Method", "role": "proposed_method", "latex": "P(h_{i}\\succ h_{j})=\\frac{\\pi_{t,h_{i}}}{\\pi_{t,h_{i}}+\\pi_{t,h_{j}}},\\qquad f_{t}(h)=a\\log\\pi_{t,h}+b_{t},\\quad a>0.", "mathml": "P⁡(hi≻hj)=πt,hiπt,hi+πt,hj,ft​(h)=a​log⁡πt,h+bt,a>0.", "context": "Population-Level Fitness. Population fitness aggregates pairwise scientific judgments into a ranking. The scorer compares every unordered pair in P0 at initialization and in the parent-offspring pool Qt at each subsequent generation. A Bradley-Terry model [5] converts the comparison outcomes into positive latent strengths πt,h and the resulting search fitness,", "headline": "", "explanation": "", "objective": "", "symbols": {}, "evidence": "", "evidence_url": "", "source_url": "" } ], "equations_status": "selected", "core_kind": "formula", "equations_fetched_at": "2026-10-09", "equations_source_url": "", "equations_source_sha256": "", "exposures_reviewed_at": "", "exposures_source_sha256": "" }