{
  "schema_version": 1,
  "builder_plan": {
    "scenario": "Start with one repeated product task that has a measurable failure or cost.",
    "change": "Read the methods and identify one change you can isolate in that workflow.",
    "baseline": "Your current implementation on the same held-out tasks.",
    "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": "Successful outcomes, response time, human rework, and total cost per successful 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": "2610.28562",
  "title": "OptScientist: Automated Discovery of Optimization Laws",
  "summary": "OptScientist replaces manual theory building and conventional symbolic regression by inferring compact symbolic optimization problems from data and wrapping the model in an LLM agent that supplies literature priors and proposes new experiments. It costs an agentic LLM pipeline (API spend or a hosted model), a custom-trained inference model, and an iterative data-collection loop rather than a single offline fit. It can fail when the true phenomenon is not well modeled as optimization, when the symbolic search overfits small datasets, or when literature priors bias discovery toward plausible but incorrect laws.",
  "rationale": "This is a scientific-discovery research framework rather than a production inference, training, or agent-engineering technique, so a constrained-infrastructure ML engineer should track it but has no production workload it directly improves.",
  "technique": "LLM-agent symbolic discovery of optimization laws",
  "family": "outro",
  "recommendation": "observar",
  "published": "2026-10-20",
  "as_of": "2026-10-21",
  "editorial_status": "indexed",
  "source_url": "https://arxiv.org/abs/2610.28562",
  "independent_implementations": 0,
  "total_implementations": 1,
  "stars": 0,
  "citations": null,
  "claims": [
    {
      "claim_id": "claim-01",
      "statement": "OptScientist replaces manual theory building and conventional symbolic regression by inferring compact symbolic optimization problems from data and wrapping the model in an LLM agent that supplies literature priors and proposes new experiments. It costs an agentic LLM pipeline (API spend or a hosted model), a custom-trained inference model, and an iterative data-collection loop rather than a single offline fit. It can fail when the true phenomenon is not well modeled as optimization, when the symbolic search overfits small datasets, or when literature priors bias discovery toward plausible but incorrect laws.",
      "basis": "inferred",
      "result": "",
      "baseline": "",
      "conditions": "",
      "source_url": "https://arxiv.org/abs/2610.28562",
      "source_page": null,
      "source_excerpt": ""
    },
    {
      "claim_id": "claim-02",
      "statement": "Recovers ground-truth laws on over 80% of the 99 OptFeynman benchmark problems and matches leading baselines on held-out prediction for human risky choice and monetary policy, with fewer parameters and tokens than LLM baselines; no multiplicative speed, memory, or cost factor is reported.",
      "basis": "inferred",
      "result": "",
      "baseline": "",
      "conditions": "",
      "source_url": "https://arxiv.org/abs/2610.28562",
      "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": ""
}