scicode-lint: Detecting Methodology Bugs in Scientific Python Code with LLM-Generated Patterns

Sergey V. Samsonau · 2026

Read scicode-lint: Detecting Methodology Bugs in Scientific Python Code with LLM-Generated Patterns by Sergey V. Samsonau (2026) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2603-17893.txt.

Description

Samsonau presents scicode-lint, a checker for method bugs in scientific Python code. Such bugs yield plausible but wrong results that normal linters miss. Frontier models draft the detection patterns at build time, while a small local model runs them. New library versions cost tokens to adapt to, not engineering hours, he argues. On Kaggle notebooks, leakage checks hit 65 percent precision at full recall. Controlled tests scored 97.7 percent across 66 patterns.

Catalog record

CREATORSergey V. Samsonau
YEAR2026
COLLECTIONPAPERS
FORMATTXT
FILE SIZE31952
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2603.17893

Browse the full library · About the Tower of Babel