AI Science Tracker
EN

An AI-assisted audit found eight maintainer-confirmed software regressions

Beihang University; National University of Singapore · 26 Oct 2026 · computer science

Discovery

An AI-assisted audit found eight maintainer-confirmed software regressions

Source check: Source checked. Not independent scientific confirmation.

Finding: An audit of 1,459 merged pull requests in four Python projects used language models to generate behavior checks. The study found nine reproducible regression candidates, of which eight were confirmed by maintainers. The result concerns a targeted software audit.

Technology: ORION; learned test generation and behavioral oracles

Publication: Preprint · Test context: prospective or independent check. Catalog categories, not an evidence grade.

AI contribution
ORION uses a learned model to infer intended software behavior and generate checks for regression candidates in existing projects.
Test and result
ORION flagged 28 requests. The authors retained and manually reviewed currently reproducible warnings before submitting nine candidates for maintainer confirmation.
Limitation / still unknown
The reviewers did not comprehensively assess unflagged changes, so recall and general defect rates are unknown. This is not a count of security vulnerabilities.
Evidence sources and annotation context

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

Evidence sources: https://arxiv.org/abs/2610.35223v1 · https://arxiv.org/html/2610.35223v1

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: v1 ORION design and §V-E real-world regression evaluation.

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 4/5 · 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 computer science 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.