AI Science Tracker
EN

An audit found errors in screened AI-extracted electro-optic values

New Jersey Institute of Technology; University of Illinois Urbana-Champaign · 25 Oct 2026 · materials

Discovery

An audit found errors in screened AI-extracted electro-optic values

Source check: Source checked. Not independent scientific confirmation.

Finding: A curated-reference audit found that 30 of 112 electro-optic records extracted with GPT-4o and accepted by automated checks were defective. Errors included incorrect tensor components, units and measurement conditions. The audit covered three selected legacy papers and does not estimate error rates for the full database.

Technology: GPT-4o; materials literature extraction

Family GPT-4: profile · models page

Publication: Preprint · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
GPT-4o extracted reported electro-optic property values and contexts from primary documents for database construction.
Test and result
The audit reviewed records already accepted by the extraction pipeline against curated source information and physical consistency checks.
Limitation / still unknown
These papers were selected as a stress test. Different formats and data complexity could yield different error rates.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://arxiv.org/abs/2610.34642v1 · https://arxiv.org/pdf/2610.34642

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: linked separately.

Review date: 21 Nov 2026. Review scope: primary v1-only PDF §§3 and 5.2, production audit and source comparison.

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 2/5 · Validation 2/5 · Novelty 2/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Open in the full catalog for filters, charts and related discoveries. Browse materials discoveries.

Related catalog entries

Links use documented relationships where available, plus shared catalog context. A shared topic or model family does not establish a scientific dependency.