Benchmarking Coding Agents Under Evolving Requirements A new arXiv paper, submitted 26 October 2026 and revised 29 October 2026, introduces FeatureLoop, a benchmark of 61 human-reviewed tasks drawn from 34 repositories across five primary programming languages and seven software forms, designed to test coding agents under evolving requirements. Evaluating 12 frontier models, the authors found the strongest model completes 72.1% of tasks when given both evolution and completion feedback, while across all models only 15.4% of observed failed stages recover within two corrective revisions. The results point to incomplete repairs, regressions and model-specific differences in feedback use and token efficiency as obstacles to reliable, low-intervention development. Computer Science Software Engineering Submitted on 26 Oct 2026 v1 https://arxiv.org/abs/2610.35488v1 , last revised 29 Oct 2026 this version, v2 Title:Benchmarking Coding Agents under Evolving Requirements View PDF https://arxiv.org/pdf/2610.35488 HTML experimental https://arxiv.org/html/2610.35488v2 Abstract:Software development with coding agents often begins with a rough request that users refine after trying the implementation. Final correctness alone does not capture the repeated guidance this process requires. We introduce FeatureLoop, a benchmark of 61 human-reviewed tasks from 34 repositories across five primary programming languages and seven software forms. Each task follows one development goal in a persistent workspace through evolution feedback, which adds or clarifies requirements, and completion feedback, which reports failures of stated requirements. Cumulative executable checks track new behavior, repair, and preservation of earlier functionality. We evaluate 12 frontier models and find that the strongest completes 72.1% of tasks with both feedback types. Across models, however, only 15.4% of observed failed stages recover within two corrective revisions. Further analysis reveals incomplete repairs, regressions, and model-specific differences in feedback use and token efficiency, identifying obstacles to reliable, low-intervention development. Submission history From: Jia Feng view email https://arxiv.org/show-email/d7b6b379/2610.35488 Mon, 26 Oct 2026 14:15:53 UTC 7,385 KB \ v1\ https://arxiv.org/abs/2610.35488v1 v2 Thu, 29 Oct 2026 03:37:42 UTC 7,385 KB Additional Features References & Citations Loading... Bibliographic and Citation Tools Bibliographic Explorer What is the Explorer? https://info.arxiv.org/labs/showcase.html arxiv-bibliographic-explorer Connected Papers What is Connected Papers? https://www.connectedpapers.com/about Litmaps What is Litmaps? https://www.litmaps.co/ scite Smart Citations What are Smart Citations? https://www.scite.ai/ Code, Data and Media Associated with this Article alphaXiv What is alphaXiv? https://alphaxiv.org/ CatalyzeX Code Finder for Papers What is CatalyzeX? https://www.catalyzex.com DagsHub What is DagsHub? https://dagshub.com/ Gotit.pub What is GotitPub? http://gotit.pub/faq Hugging Face What is Huggingface? https://huggingface.co/huggingface ScienceCast What is ScienceCast? https://sciencecast.org/welcome Demos Recommenders and Search Tools Influence Flower What are Influence Flowers? https://influencemap.cmlab.dev/ CORE Recommender What is CORE? https://core.ac.uk/services/recommender arXivLabs: experimental projects with community collaborators arXivLabs is a framework that allows collaborators to develop and share new arXiv features directly on our website. Both individuals and organizations that work with arXivLabs have embraced and accepted our values of openness, community, excellence, and user data privacy. arXiv is committed to these values and only works with partners that adhere to them. Have an idea for a project that will add value for arXiv's community? Learn more about arXivLabs https://info.arxiv.org/labs/index.html .