Preprint finds targeted edge tests reduce faulty code repair releases

Published Friday, November 6, 2026 · 8:52 AM EST

A preregistered preprint reports that adding one edge-case input derived from a release note reduced the share of wrong code repairs passing a release check from 0.410 to 0.065. The study covered 73 tasks spanning 60 documented runtime or API changes and executed candidates on real runtimes against withheld contracts. The result reached 0.061 on 15 unseen clusters. Later stages added more candidate fragments and hard distractors that matched legacy output on typical inputs. Even then, 0.168 of hard wrong programs passed the edge-input check. No model-written code was executed, and disputes over contract labels remain open.

Sources

cs.AI updates on arXiv.org

Published by Tech & Business, a media brand covering technology and business. This story was sourced from cs.AI updates on arXiv.org and reviewed by the T&B editorial agent team.

Back to Newswire

Keep reading

Full wire

Products

Fox closes $22 billion Roku acquisition

Fox closed its $22 billion acquisition of Roku on Friday, the companies announced, giving it control of a streaming platform reaching more than 100 million US households. Roku founder Anthony Wood now serves on Fox's board.

Products

Leidos reports $5.5 billion in bookings and raises earnings outlook

Leidos reported $5.5 billion in third-quarter net bookings and a $49.6 billion backlog, including $11.1 billion in funded work. Revenue reached $4.63 billion, up 4% from the prior-year period.

Capital Robotics

Physicl raises $10 million and releases robot simulation tools

Physicl has raised $10 million in seed funding from Inovia Capital and is releasing Studio, a tool for generating synthetic robot training data, TechCrunch reports.

Security

Gnosis Pay discloses customer data breach and adds payment checks

Gnosis Pay disclosed a breach of its consumer card infrastructure that exposed personal and payment-related customer information, Coingape reported.

Capital

Aave Labs acquires Avon team and lending technology

Aave Labs has acquired Avon's team and technology to expand its lending services and develop more fixed-rate borrowing options, KuCoin reported. The assets include fixed-rate and floating-rate lending protocols, an onchain order book, domains and related assets.

Security

Veeam distributes fixes for critical Linux kernel vulnerability

Veeam said on Friday that some of its products are affected by CVE-2026-98432, a critical Linux kernel vulnerability, BleepingComputer reported. The flaw allows remote unauthenticated attackers to execute commands as root on servers running Multipath TCP listeners.