{"resourceId":"verified-agentic-ingestion-scientific-tables-agents-propose-code-verifies-humans-decide-20261108","versions":[{"version":"external-14db3a30ff9f8f73bd486d188ff0268da39a45f9ee0e9fc9a012be665edfa1af","resource":{"id":"verified-agentic-ingestion-scientific-tables-agents-propose-code-verifies-humans-decide-20261108","title":"Agents Propose, Code Verifies, Humans Decide: Verified Agentic Ingestion of Heterogeneous Scientific Tables into Relational Databases","organization":"Colorado State University and The Ohio State University","sector":"Research data management (biomedical dataset ingestion)","geography":"United States (authors' institutions); four public biomedical datasets","publishedAt":"8 November 2026 (arXiv 2611.11045v1, cs.DB)","publicationDate":"2026-11-08","eventDate":null,"sourceName":"arXiv","sourceLabel":"Rankothge, B., Hodgkins, S., Rich, V., and Kirby, M., \"Agents Propose, Code Verifies, Humans Decide,\" arXiv:2611.11045, 8 Nov 2026","sourceUrl":"https://arxiv.org/abs/2611.11045","evidenceClass":"academic-research","outcomeClass":"effective","topics":["knowledge-work","data-security","governance-procurement"],"finding":"The authors describe an ingestion pipeline that turns messy public scientific tables (wide matrices, long tables, per-cohort files) into a PostgreSQL database under one rule: LLM agents propose typed schemas from compact evidence about the files, a deterministic verifier with 39 check types accepts or rejects each proposal in a structure loop and a meaning loop, and a human approves the schema and the load. On four public biomedical datasets, one held out from development, 9 of 20 first proposals failed a check and every failure was caught; the loops repaired eight and a one-line reviewer edit fixed the ninth. All 45.1 million values loaded with zero reconciliation gap, structure discovery took 4-7 LLM calls and data preparation at most one more, at under $0.50 per dataset. In a separate test, each of nine planted data errors was quarantined with a reason or recorded in an audit log. The paper illustrates silent failure with GSE73072, where keying on a subject column alone would keep 30 subjects instead of 151. Included now because it was submitted 8 Nov, appeared in the 10 Nov arXiv listings and was not archived.","sledRelevance":"Relevant to university data-management, library and research-computing groups that help investigators reuse public datasets: it shows a controlled pattern for letting agents do the structural work while code and a reviewer own correctness, with measured cost on four datasets.","evidence":"arXiv preprint reporting results on four public biomedical datasets (one held out from development), 45.1 million values, per-dataset call counts and cost, and a planted-error test with nine errors. No comparison against an unconstrained coding agent. Not peer reviewed.","architectureImplications":"Narrow agents receive measured evidence, never the full data or database access; a verifier compares typed proposals with evidence; code compiles DDL, reshapes data, quarantines unplaceable rows, records provenance and reconciles every staged row; two human checkpoints precede schema creation and load.","governanceImplications":"Human approval of the schema and of the load, provenance records and quarantine reasons give an audit trail. Reviewers remain accountable; the verifier does not check where descriptive columns belong, which varied across runs on one dataset.","securityPrivacyImplications":"Agents see compact evidence rather than raw data, which limits exposure to the model provider; the datasets studied were public, so controlled or identifiable research data would need separate hosting and access decisions that the paper does not address.","caveats":"Preprint; four biomedical datasets and one model family; no baseline against an unconstrained agent (the subject-ID case is an illustration, not a failure rate); kept values on repeated keys depend on file order; large wide matrices enlarge prompts; costs are LLM-call costs only, not engineering or reviewer time.","streamIds":["research"],"roles":{"sales":"Framing point: researchers lose time reconnecting messy public datasets, and this preprint shows agents can help when they only propose. On four biomedical datasets, 9 of 20 first schema proposals failed a deterministic check, all were caught, and 45.1 million values loaded with no reconciliation gap for under $0.50 per dataset in LLM calls. Use it with research data services, libraries and bioinformatics cores. Position a bounded ingestion pilot on a few local datasets, and avoid claiming staff-time savings or applicability beyond the biomedical sets tested.","engineering":"Framing point: the design principle is separation of proposal from execution. Agents return typed proposals from compact file evidence; a verifier with 39 checks tests structure and meaning; code builds PostgreSQL DDL, loads in batches, quarantines rows and reconciles counts; humans approve schema and load. Integration needs a PostgreSQL target, file profiling, and an LLM endpoint. Known gaps: placement of descriptive columns is unchecked, first-row-wins on duplicate keys depends on file order, and wide matrices strain prompts. Test on your own file shapes.","delivery":"Framing point: the workflow only works if reviewers actually approve each schema and load. Plan dataset selection, reviewer roles, check-type tuning to local data conventions, and a quarantine review routine. Assign an owner for the verifier rules and the audit log. The paper gives no staffing or elapsed-time data, so measure reviewer hours and rework in a pilot, and keep the human checkpoints even if first-pass failure rates fall."},"roleDetails":{"sales":{"customerProblem":"Reanalysis of public scientific datasets requires reconnecting inconsistent files, and automated loaders can succeed silently with wrong keys.","stakeholders":["Research data services lead","Bioinformatics or data core director","Library data curators","Principal investigators reusing public cohorts"],"discoveryQuestions":["Which public or collaborator datasets are reloaded most often?","How are ingestion errors detected today?","Who approves a schema before load?","Do any datasets carry controlled-access restrictions?"],"valueHypothesis":"Agent-proposed schemas gated by deterministic checks and human approval may reduce manual ingestion work while preserving an audit trail; the paper measures LLM calls and cost, not staff time.","potentialEngagement":"A pilot ingesting two or three local datasets with the propose-verify-approve pattern, comparing errors caught, reviewer effort and reconciliation results with the current process.","unsupportedClaims":["Staff-time or total cost savings","Performance on non-biomedical or controlled-access data","Superiority over other coding agents","That verification removes the need for human approval"]},"engineering":{"fit":"Good for table-based scientific data with identifiable keys; evidence covers four biomedical datasets and one model family.","architectureIntegration":"Profile files into compact evidence, call narrow proposal agents, run a 39-check verifier in structure and meaning loops, compile DDL, load via batched COPY, quarantine and reconcile.","prerequisites":["PostgreSQL target","File profiling that produces citable evidence","LLM endpoint","Reviewer interface for schema and load approval"],"constraints":["Kept value on duplicate keys depends on file order","Descriptive-column placement is unverified","Wide matrices enlarge prompts","No unconstrained-agent baseline"],"security":"Only compact evidence goes to the model, but controlled or identifiable data needs its own hosting, access and retention review.","proofOfValue":["Run local datasets through the pipeline and record first-proposal check failures","Plant known key and encoding errors and confirm each is quarantined or logged","Reconcile staged and loaded row counts"]},"delivery":{"work":"Select pilot datasets, adapt check types to local conventions, set reviewer checkpoints and run the load with reconciliation.","dependencies":["Reviewer time","Database environment","Dataset documentation"],"ownership":"The research data service owns the pipeline and verifier rules; the dataset's investigator approves the schema and load.","skillsAdoption":"Data stewards learn to review typed schema proposals and quarantine reasons rather than write loaders by hand.","governanceCheckpoints":["Schema approval before any table is created","Load approval before rows are inserted","Review of quarantined rows"],"acceptanceCriteria":["Proposed: every load reconciles to zero gap","Proposed: planted errors are quarantined or logged","Proposed: reviewer hours per dataset recorded"],"risks":["Rubber-stamp approvals","Verifier rules missing local modeling issues","Unreviewed quarantine backlog","Prompt growth on wide matrices"]}},"retrievedAt":"2026-11-10T22:05:45-06:00","enrichedAt":"2026-11-10T22:05:45-06:00","enrichmentBasis":"retrieved source","accessibilityWorkforceImplications":"Moves data-wrangling effort toward review and rule-writing; the paper offers no evidence on workload or training needs.","procurementImplications":"When evaluating agentic data tools, ask for deterministic verification, provenance and reconciliation features and measured first-pass failure rates, not only success rates.","operatingModelImplications":"Supports a shared research-data service where agents draft schemas, code enforces rules and named humans accept each load.","peerReview":"preprint","sourceVerification":{"openedUrl":"https://arxiv.org/abs/2611.11045","referenceExcerpt":"9 of 20 first proposals failed a check and every failure was caught","promptVersion":"sled-research-v4.0","model":null,"basis":"agent-reported inspection"}}}]}