Xiaonan Xu, Wenjing Wu · 2026 · arXiv
DOI: 10.48550/arXiv.2610.30571 · open-access PDF
Tujuan: Agen pengkodean mengikuti file instruksi proyek, seperti CLAUDE.md atau AGENTS.md, bahkan ketika tidak setuju dengan kode yang ada dan bahkan ketika mereka melihat ketidaksetujuan. Kami mengukur perilaku ini dan menguji apakah satu kalimat dalam file instruksi pengguna membuat agen mengikuti kode sebagai gantinya. Metode: Kami membangun 40 proyek Python di mana file instruksi meminta satu konvensi sementara kode yang ada menggunakan yang lain, misalnya cetak bukan logging. Claude Code dan Codex CLI menambahkan fitur ke setiap proyek, dengan dan tanpa kalimat dalam file instruksi tingkat pengguna
Machine translation (Meta M2M100); the English original follows.
Purpose: Coding agents follow a project's instruction file, such as CLAUDE.md or AGENTS.md, even when it disagrees with the existing code and even when they notice the disagreement. We measure this behaviour and test whether one sentence in the user's own instruction file makes agents follow the code instead. Methods: We built 40 Python projects in which the instruction file asks for one convention while the existing code uses another, for example print instead of logging. Claude Code and Codex CLI added a feature to each project, with and without a sentence in the user-level instruction file
Read in: English · Spanish · Chinese · Hindi · Arabic · Portuguese · French · German · Japanese · Russian · Korean · Indonesian · Vietnamese
GET https://store.agentexchange.work/gw/papers/verify?title=...&year=... (0.003 USD), fetch this record as JSON with /gw/papers/paper?id=2610.30571, or list what it cites with /gw/papers/references?id=.... Pay per call in USDC (x402) or with a 19 USD card key for 1,000 calls. Catalogue · llms.txtSources: OpenAlex (CC0), arXiv, Crossref. Agent Exchange