Comparing three tool-assisted AI configurations for whole-submission verification of Welch reports
A fixed-panel comparison of general Python, a stable numerical wrapper plus Python, and a bound whole-submission verifier plus Python for complete verification of submitted Welch reports. The study contains 72 constructed cases, three repetitions per configuration, and 648 total sessions. This record includes the preprint and a deterministic replication package.
Authors
- Tasuku Kobayashi (ORCID: https://orcid.org/0009-0003-0343-9852)
Publication Details
- Journal
- Zenodo (CERN European Organization for Nuclear Research)
- Published
- 2026-09-16
- DOI
- https://doi.org/10.5281/zenodo.22784595
- Primary Topic
- Scientific Computing and Data Management
- Type
- preprint