{"ID":2843732,"CreatedAt":"2026-06-01T04:54:23.091178241Z","UpdatedAt":"2026-06-01T04:54:23.091178241Z","DeletedAt":null,"paper_url":"https://arxiv.org/abs/2511.06701","arxiv_id":"2511.06701","title":"Structural Enforcement of Statistical Rigor in AI-Driven Discovery: A Functional Architecture","abstract":"AI-Scientist systems that use large language models to automate research risk generating spurious discoveries through uncontrolled multiple testing. We present a functional architecture that enforces statistical rigor at two levels: a Haskell embedded DSL (the Research monad) that makes it impossible to test a hypothesis without updating the error budget, and a declarative scaffolding technique that structurally prevents data leakage across the boundary into LLM-generated code. We ground these guarantees in a machine-checked Lean 4 formalization of the LORD++ online FDR control theorem (855 lines, zero sorry), which identifies four sufficient conditions for FDR control. Three are structural conditions -- about information flow, data separation, and test validity -- enforced by the architecture's type system and scaffolding. The fourth is an arithmetic condition: a budget invariant requiring that a wealth process remain non-negative under floating-point computation. We verify this condition over IEEE 754 doubles using SPARK/Ada, whose GNATprove toolchain statically confirms that no rounding sequence can violate the invariant and whose flow analysis independently confirms the predictability condition. The resulting verification chain -- from real-analysis proof to floating-point implementation -- is, to our knowledge, the first for any online FDR control procedure. Monte Carlo simulation (N=2000 hypotheses) and an end-to-end case study confirm that the monadic implementation controls FDR at 1.1% against a 5% target, while a naive approach inflates to 41%.","short_abstract":"AI-Scientist systems that use large language models to automate research risk generating spurious discoveries through uncontrolled multiple testing. We present a functional architecture that enforces statistical rigor at two levels: a Haskell embedded DSL (the Research monad) that makes it impossible to test a hypothes...","url_abs":"https://arxiv.org/abs/2511.06701","url_pdf":"https://arxiv.org/pdf/2511.06701v2","authors":"[\"Karen Sargsyan\"]","published":"2025-11-10T04:42:30Z","proceeding":"cs.SE","tasks":"[\"cs.SE\",\"cs.AI\"]","methods":"[\"Large Language Model\",\"Language Model\"]","has_code":false}