{"status":"ok","message-type":"work","message-version":"1.0.0","message":{"indexed":{"date-parts":[[2026,10,7]],"date-time":"2026-10-07T22:49:47Z","timestamp":1791413387213,"version":"4.3.3"},"reference-count":41,"publisher":"Association for Computing Machinery (ACM)","issue":"10","license":[{"start":{"date-parts":[[2026,9,25]],"date-time":"2026-09-25T00:00:00Z","timestamp":1790294400000},"content-version":"vor","delay-in-days":0,"URL":"https:\/\/creativecommons.org\/licenses\/by\/4.0\/"}],"content-domain":{"domain":["dl.acm.org"],"crossmark-restriction":true},"short-container-title":["Commun. ACM"],"published-print":{"date-parts":[[2026,10,1]]},"abstract":"<jats:p>Self-healing systems have long been a focus of research, aiming to enable software to recover from unexpected runtime errors without human intervention. Traditional approaches rely on predefined heuristic rules, such as reusing error handlers or rolling back to checkpoints, but these methods struggle to adapt to the diverse range of runtime errors. The emergence of large language models offers a new opportunity to address this challenge. Leveraging their ability to understand and generate code and natural language, we propose using LLMs to dynamically generate error-handling strategies in real time, tailored to specific runtime contexts such as error messages and program states.<\/jats:p>\n                  <jats:p>\n                    We demonstrate the feasibility of this approach by designing such a framework,\n                    <jats:sc>Healer<\/jats:sc>\n                    , and empirically showing that it can handle runtime errors with a high success rate. When an unanticipated runtime error occurs,\n                    <jats:sc>Healer<\/jats:sc>\n                    leverages its internal LLM to generate bespoke error-handling code. The LLM is prompted with runtime information, including the error message, error location, and current program state. The generated healing code is then executed to produce a corrected program state, allowing the program to continue execution with minimal disruption. We evaluate\n                    <jats:sc>Healer<\/jats:sc>\n                    across four code datasets and three state-of-the-art LLMs (GPT-3.5, GPT-4, and CodeQwen-7B), where GPT-4 can successfully recover from 72.8% of runtime errors, underscoring the promise of LLMs in this domain. Despite these promising results, challenges remain, particularly regarding the trustworthiness of LLM-generated code and its integration into existing systems. We mention potential solutions, such as safety checks and\n                    <jats:sc>Healer<\/jats:sc>\n                    -aware programming, to mitigate risks and ensure reliable operation. This work represents the first step toward agentic runtime healing, paving the way for more adaptive, resilient, and self-healing software systems.\n                  <\/jats:p>","DOI":"10.1145\/3800691","type":"journal-article","created":{"date-parts":[[2026,9,25]],"date-time":"2026-09-25T14:44:08Z","timestamp":1790347448000},"page":"88-96","update-policy":"https:\/\/doi.org\/10.1145\/crossmark-policy","source":"Crossref","is-referenced-by-count":0,"title":["Toward Agentic Runtime Healing"],"prefix":"10.1145","volume":"69","author":[{"ORCID":"https:\/\/orcid.org\/0000-0001-5393-7858","authenticated-orcid":false,"given":"Zhensu","family":"Sun","sequence":"first","affiliation":[{"name":"Singapore Management University, Singapore, Singapore, Singapore"}],"role":[{"vocabulary":"crossref","role":"author"}]},{"ORCID":"https:\/\/orcid.org\/0009-0007-2520-8309","authenticated-orcid":false,"given":"Haotian","family":"Zhu","sequence":"additional","affiliation":[{"name":"Singapore Management University, Singapore, Singapore, Singapore"}],"role":[{"vocabulary":"crossref","role":"author"}]},{"ORCID":"https:\/\/orcid.org\/0000-0002-1006-8493","authenticated-orcid":false,"given":"Bowen","family":"Xu","sequence":"additional","affiliation":[{"name":"North Carolina State University at Raleigh, Raleigh, North Carolina, United States"}],"role":[{"vocabulary":"crossref","role":"author"}]},{"ORCID":"https:\/\/orcid.org\/0000-0003-3728-9541","authenticated-orcid":false,"given":"Xiaoning","family":"Du","sequence":"additional","affiliation":[{"name":"Monash University, Melbourne, Victoria, Australia"}],"role":[{"vocabulary":"crossref","role":"author"}]},{"ORCID":"https:\/\/orcid.org\/0000-0003-2990-1614","authenticated-orcid":false,"given":"Li","family":"Li","sequence":"additional","affiliation":[{"name":"Beihang University, Beijing, Beijing, China"}],"role":[{"vocabulary":"crossref","role":"author"}]},{"ORCID":"https:\/\/orcid.org\/0000-0002-4367-7201","authenticated-orcid":false,"given":"David","family":"Lo","sequence":"additional","affiliation":[{"name":"Singapore Management University, Singapore, Singapore, Singapore"}],"role":[{"vocabulary":"crossref","role":"author"}]}],"member":"320","published-online":{"date-parts":[[2026,9,25]]},"reference":[{"key":"e_1_3_1_2_2","unstructured":"Achiam J. et al. GPT-4 technical report. arXiv preprint (2023) \u00a0arXiv:2303.08774."},{"key":"e_1_3_1_3_2","unstructured":"ajalt. fuckitpy: The Python error steamroller\u00a0(2026); https:\/\/github.com\/ajalt\/fuckitpy"},{"key":"e_1_3_1_4_2","unstructured":"AlphaCode Team Google DeepMind. AlphaCode 2 technical report\u00a0(2023); https:\/\/storage.googleapis.com\/deepmind-media\/AlphaCode2\/AlphaCode2_Tech_Report.pdf\u00a0"},{"key":"e_1_3_1_5_2","doi-asserted-by":"crossref","unstructured":"Candea G. et al. JAGR: An autonomous self-recovering application server. In 2003\u00a0Autonomic\u00a0Computing\u00a0Workshop. IEEE\u00a0(2003) 168\u2013177.","DOI":"10.1109\/ACW.2003.1210217"},{"key":"e_1_3_1_6_2","doi-asserted-by":"crossref","unstructured":"Carzaniga A. et al. Automatic recovery from runtime failures. In 2013 35th Intern. Conf. on Software Engineering. IEEE\u00a0(2013) 782\u2013791.","DOI":"10.1109\/ICSE.2013.6606624"},{"key":"e_1_3_1_7_2","doi-asserted-by":"crossref","unstructured":"Carzaniga A. Gorla A. Perino N. and Pezz\u00e8 M. Automatic workarounds for web applications. In Proceedings\u00a0of\u00a0the\u00a0Eighteenth\u00a0ACM\u00a0SIGSOFT\u00a0Intern.\u00a0Symp.\u00a0on\u00a0Foundations\u00a0of\u00a0Software\u00a0Engineering. ACM (2010) 237\u2013246.","DOI":"10.1145\/1882291.1882327"},{"key":"e_1_3_1_8_2","doi-asserted-by":"publisher","DOI":"10.1145\/2522920.2522923"},{"key":"e_1_3_1_9_2","unstructured":"\u00a0Common Weakness Enumeration. CWE-248: Uncaught exception (4.14); https:\/\/cwe.mitre.org\/data\/definitions\/248.html\u00a0"},{"key":"e_1_3_1_10_2","unstructured":"Diamond M. fuckitjs: The original Javascript error steamroller\u00a0(2026); https:\/\/github.com\/mattdiamond\/fuckitjs\u00a0"},{"key":"e_1_3_1_11_2","unstructured":"EvalPlus. EvalPlus Releases\u00a0(2024); https:\/\/github.com\/evalplus\/evalplus\/releases\u00a0"},{"key":"e_1_3_1_12_2","unstructured":"Gerganov G. llama.cpp: Port of Facebook\u2019s LLaMA model in C\/C++\u00a0(2023); https:\/\/github.com\/ggerganov\/llama.cpp\u00a0"},{"key":"e_1_3_1_13_2","doi-asserted-by":"publisher","DOI":"10.1016\/j.dss.2006.06.011"},{"key":"e_1_3_1_14_2","doi-asserted-by":"crossref","unstructured":"Gu T. et al. Automatic runtime recovery via error handler synthesis. In Proceedings of the 31st IEEE\/ACM Intern. Conf. on Automated Software Engineering. ACM\u00a0 (2016) 684\u2013695.","DOI":"10.1145\/2970276.2970360"},{"key":"e_1_3_1_15_2","unstructured":"Hou X. et al. Large language models for software engineering: A systematic literature review. arXiv preprint (2023) \u00a0arXiv:2308.10620.\u00a0"},{"key":"e_1_3_1_16_2","unstructured":"Hu E.J. et al. Lora: Low-rank adaptation of large language models. arXiv preprint (2021) \u00a0arXiv:2106.09685."},{"key":"e_1_3_1_17_2","unstructured":"IBM. Project CodeNet\u00a0(2024); https:\/\/github.com\/IBM\/Project_CodeNet\u00a0"},{"key":"e_1_3_1_18_2","unstructured":"Jain N. et al. LiveCodeBench: Holistic and contamination free evaluation of large language models for code. The\u00a0Thirteenth\u00a0Intern.\u00a0Conf.\u00a0On\u00a0Learning\u00a0Representations\u00a0(2025)."},{"key":"e_1_3_1_19_2","unstructured":"Jiang A.Q. et al. Mistral 7B. arXiv preprint\u00a0(2023)"},{"key":"e_1_3_1_20_2","doi-asserted-by":"crossref","unstructured":"Kochhar P.S. et al. Understanding the test automation culture of app developers. In 2015 IEEE 8th Intern. Conf. on Software Testing Verification and Validation. IEEE\u00a0(2015) 1\u201310.","DOI":"10.1109\/ICST.2015.7102609"},{"key":"e_1_3_1_21_2","unstructured":"Li R. et al. StarCoder: May the source be with you! arXiv preprint (2023) arXiv:2305.06161."},{"key":"e_1_3_1_22_2","doi-asserted-by":"publisher","DOI":"10.1016\/j.neucom.2021.07.045"},{"key":"e_1_3_1_23_2","article-title":"Is your code generated by ChatGPT really correct? Rigorous evaluation of large language models for code generation","volume":"36","author":"Liu J.","year":"2024","unstructured":"Liu, J., Xia, C.S., Wang, Y., and Zhang, L. Is your code generated by ChatGPT really correct? Rigorous evaluation of large language models for code generation. Advances\u00a0in\u00a0Neural\u00a0Information\u00a0Processing\u00a0Systems 36 (2024).","journal-title":"Advances\u00a0in\u00a0Neural\u00a0Information\u00a0Processing\u00a0Systems"},{"key":"e_1_3_1_24_2","doi-asserted-by":"crossref","unstructured":"Long F. et al. Automatic input rectification. In 2012 34th Intern. Conf. on Software Engineering. IEEE\u00a0(2012) 80\u201390.","DOI":"10.1109\/ICSE.2012.6227204"},{"key":"e_1_3_1_25_2","unstructured":"OpenAI. GPT-3.5 Turbo model documentation\u00a0(2024); https:\/\/platform.openai.com\/docs\/models\/gpt-3-5-turbo"},{"key":"e_1_3_1_26_2","unstructured":"OpenAI. Prompt engineering: Use delimiters to clearly indicate distinct parts of the input\u00a0(2024); https:\/\/platform.openai.com\/docs\/guides\/prompt-engineering\/tactic-use-delimiters-to-clearly-indicate-distinct-parts-of-the-input"},{"key":"e_1_3_1_27_2","unstructured":"OpenAI. openai\/gpt-oss-20b - Hugging Face\u00a0(2025); https:\/\/huggingface.co\/openai\/gpt-oss-20b\u00a0"},{"key":"e_1_3_1_28_2","doi-asserted-by":"crossref","unstructured":"Pearce H. et al. Asleep at the keyboard? Assessing the security of github copilot\u2019s code contributions. In 2022\u00a0IEEE\u00a0Symp.\u00a0On\u00a0Security\u00a0and\u00a0Privacy. IEEE\u00a0(2022) 754\u2013768.","DOI":"10.1109\/SP46214.2022.9833571"},{"key":"e_1_3_1_29_2","doi-asserted-by":"publisher","DOI":"10.1007\/s00607-010-0107-y"},{"key":"e_1_3_1_30_2","unstructured":"Rinard M.C. et al. Enhancing server availability and security through failure-oblivious computing. In Proceedings of the 6th conf. on Symp. on Operating Systems Design & Implementation - Volume 6. USENIX (2004) 21\u201321."},{"key":"e_1_3_1_31_2","unstructured":"Roziere B. et al. Code Llama: Open foundation models for code. arXiv preprint (2023) arXiv:2308.12950."},{"key":"e_1_3_1_32_2","doi-asserted-by":"crossref","unstructured":"Shi J. et al. Compressing pre-trained models of code into 3 mb. In Proceedings of the 37th IEEE\/ACM Intern. Conf. on Automated Software Engineering. IEEE (2022) 1\u201312.","DOI":"10.1145\/3551349.3556964"},{"key":"e_1_3_1_33_2","unstructured":"sosreport. sosreport\/sos: A unified tool for collecting system logs and other debug information\u00a0(2024); https:\/\/github.com\/sos-project\/sos"},{"key":"e_1_3_1_34_2","doi-asserted-by":"crossref","unstructured":"Souza B. and Pradel M. LExecutor: Learning-guided execution. In Proceedings of the 31st ACM Joint European Software Engineering Conf. and Symp. on the Foundations of Software Engineering. ACM (2023) 1522\u20131534.","DOI":"10.1145\/3611643.3616254"},{"key":"e_1_3_1_35_2","doi-asserted-by":"crossref","unstructured":"Tian R. et al. Debugbench: Evaluating debugging capability of large language models. arXiv preprint (2024) arXiv:2401.04621.","DOI":"10.18653\/v1\/2024.findings-acl.247"},{"key":"e_1_3_1_36_2","unstructured":"Wang X. et al. Openhands: An open platform for AI software developers as generalist agents. arXiv preprint (2024) arXiv:2407.16741."},{"key":"e_1_3_1_37_2","doi-asserted-by":"crossref","unstructured":"Wang Y. et al. CodeT5+: Open code large language models for code understanding and generation. arXiv preprint (2023) \u00a0arXiv:2305.07922.","DOI":"10.18653\/v1\/2023.emnlp-main.68"},{"key":"e_1_3_1_38_2","doi-asserted-by":"crossref","unstructured":"Weimer W. and Necula G.C. Finding and preventing run-time error handling mistakes. In Proceedings of the 19th annual ACM SIGPLAN Conf. on Object-oriented programming systems languages and applications. ACM (2004) 419\u2013431.","DOI":"10.1145\/1028976.1029011"},{"key":"e_1_3_1_39_2","doi-asserted-by":"crossref","unstructured":"Weyns D. et al. Preliminary results of a survey on the use of self-adaptation in industry. In Proceedings of the 17th Symp. on Software Engineering for Adaptive and Self-Managing Systems. ACM\u00a0(2022) 70\u201376.","DOI":"10.1145\/3524844.3528077"},{"key":"e_1_3_1_40_2","unstructured":"Yang Z. et al. Robustness security privacy explainability efficiency and usability of large language models for code. arXiv preprint (2024) \u00a0arXiv:2403.07506."},{"key":"e_1_3_1_41_2","doi-asserted-by":"crossref","unstructured":"Zhang F. et al. RepoCoder: Repository-level code completion through iterative retrieval and generation. In Proceedings\u00a0of\u00a0the\u00a02023\u00a0Conf.\u00a0On\u00a0Empirical\u00a0Methods\u00a0in\u00a0Natural\u00a0Language\u00a0Processing H. Bouamor J. Pino and K. Bali (Eds.). Association for Computational Linguistics\u00a0(2023) 2471\u20132484.","DOI":"10.18653\/v1\/2023.emnlp-main.151"},{"key":"e_1_3_1_42_2","doi-asserted-by":"crossref","unstructured":"Zhou J. Zhang H. and Lo D. Where should the bugs be fixed? more accurate information retrieval-based bug localization based on bug reports. In 2012 34th Intern. Conf. on Software Engineering. IEEE\u00a0(2012) 14\u201324.","DOI":"10.1109\/ICSE.2012.6227210"}],"container-title":["Communications of the ACM"],"original-title":[],"language":"en","link":[{"URL":"https:\/\/dl.acm.org\/doi\/pdf\/10.1145\/3800691","content-type":"unspecified","content-version":"vor","intended-application":"similarity-checking"}],"deposited":{"date-parts":[[2026,10,7]],"date-time":"2026-10-07T22:00:30Z","timestamp":1791410430000},"score":1,"resource":{"primary":{"URL":"https:\/\/dl.acm.org\/doi\/10.1145\/3800691"}},"subtitle":[],"short-title":[],"issued":{"date-parts":[[2026,9,25]]},"references-count":41,"journal-issue":{"issue":"10","published-print":{"date-parts":[[2026,10,1]]}},"alternative-id":["10.1145\/3800691"],"URL":"https:\/\/doi.org\/10.1145\/3800691","relation":{},"ISSN":["0001-0782","1557-7317"],"issn-type":[{"value":"0001-0782","type":"print"},{"value":"1557-7317","type":"electronic"}],"subject":[],"published":{"date-parts":[[2026,9,25]]},"assertion":[{"value":"2026-09-25","order":3,"name":"published","label":"Published","group":{"name":"publication_history","label":"Publication History"}}]}}