Newest papers

Teaching Code Agents to Ask Without Leaking Private Code

Wenjia Jiang, Jieke Shi, Zhensu Sun, Chengran Yang, Shidong Pan, Peixin Zhang, Yifan Jia, James Hoang, Xiwei Xu, Zhenchang Xing, David Lo, Zhou Yang · 2026 · arXiv (Cornell University)

DOI: 10.48550/arxiv.2610.24824 · cited by 0 works (OpenAlex) · open-access PDF

German

Lokale codierende Agenten können ihre Reichweite erweitern, indem sie ein stärkeres Cloud-Modell konsultieren. Solche Konsultationen zeigen jedoch oft die privaten Identifikatoren, Dateiwege und Quellcode des Repositoriums. Hostseitiges Masken versteckt erkannte Identifikatoren, lässt aber den umgebenden Code und die Ausdrucksweise des Agenten unverändert, während das Training für weniger Offenlegung die Konsultation unterdrückt oder Fragen zu vage hervorbringt, um zu antworten. Wir schlagen die nutz-verifizierte Abstraktion (UVA) vor, die den Agenten lehrt, private Entitäten nach ihrer Rolle und Art zu beschreiben, während die technischen Details eines Experten gebraucht werden. Von den gelöschten Reparaturpfaden überarbeiten wir jede Frage in diese Form und behalten

Machine translation (Meta M2M100); the English original follows.

Abstract

Local coding agents can extend their reach by consulting a stronger cloud model. Such consultations, however, often expose the repository's private identifiers, file paths, and source code. Host-side masking hides recognized identifiers but leaves the surrounding code and the agent's phrasing unchanged, while training for less disclosure risks suppressing consultation or producing questions too vague to answer. We propose utility-verified abstraction (UVA), which teaches the agent to describe private entities by their role and type while retaining the technical details an expert needs. From solved repair trajectories, we rewrite each question into this form and keep the rewrite only if it discloses fewer private entities and the repair, replayed from the same execution state with the repair trajectories. Across two benchmarks, the trained agent consults more often than before training. On repository-level issue repair, it solves 37.6% of issues, against 25.3% for the same model working alone, a gain from the training as a whole rather than from consultation alone. Its questions disclose 3.7-fold fewer private entities and 1.4-fold less verbatim code than the untrained agent's questions to the same expert.

Read in: English · Spanish · Chinese · Hindi · Arabic · Portuguese · French · German · Japanese · Russian · Korean · Indonesian · Vietnamese

Topics: Computer science, Identifier, Coding (social sciences), Ask price, Computer security, Code (set theory)

Sources: OpenAlex (CC0), arXiv, Crossref. Agent Exchange