Self-Evolving LLM Agents for Automated Software Engineering Through Reinforcement Learning and Execution Feedback
Keywords:
large language models; software engineering; reinforcement learning; execution feedback; autonomous agents; governance; socio-technical systems; sustainabilityAbstract
The convergence of large language models, autonomous agents, and reinforcement learning has created new possibilities for automated software engineering. This paper examines self-evolving LLM agents that improve their software engineering behavior through execution feedback, tool interaction, and reinforcement learning signals. Unlike static code assistants, self-evolving agents maintain memory, plan across tasks, invoke compilers, test suites, debuggers, version control, and deployment pipelines, and update their policies from observed outcomes. The paper develops a system-level account of these agents by analyzing architecture, reward design, execution infrastructure, lifecycle integration, robustness, governance, fairness, and sustainability. It argues that execution feedback provides a powerful but incomplete learning signal: it can verify functional correctness, expose runtime regressions, and guide iterative repair, yet it cannot fully capture maintainability, security, ethical acceptability, or long-term architectural fitness. Reinforcement learning with execution feedback therefore requires hybrid reward structures that combine automated tests, static analysis, human preferences, and organizational constraints. The paper further discusses the trade-offs among autonomy, safety, cost, latency, and auditability. It proposes that self-evolving software agents should be deployed as governed participants in socio-technical infrastructures rather than as unaccountable replacements for engineering judgment. Evaluation must move beyond benchmark pass rates toward longitudinal measures of reliability, fairness, maintainability, incident response, and human oversight. The conclusion outlines a research agenda for reproducible, sustainable, and policy-aware agentic software engineering.
References
1. Vaswani, A., Shazeer, N., Parmar, N., Uszkoreit, J., Jones, L., Gomez, A. N., Kaiser, Ł., & Polosukhin, I. (2017). Attention is all you need. Advances in Neural Information Processing Systems, 30, 5998–6008.
2. Brown, T. B., Mann, B., Ryder, N., Subbiah, M., Kaplan, J., Dhariwal, P., ... Amodei, D. (2020). Language models are few-shot learners. Advances in Neural Information Processing Systems, 33, 1877–1901.
3. Chen, M., Tworek, J., Jun, H., Yuan, Q., Pinto, H. P. d. O., Kaplan, J., ... Zaremba, W. (2021). Evaluating large language models trained on code. arXiv preprint arXiv:2107.03374.
4. Austin, J., Odena, A., Nye, M., Bosma, M., Michalewski, H., Dohan, D., ... Sutton, C. (2021). Program synthesis with large language models. arXiv preprint arXiv:2108.07732.
5. Li, Y., Choi, D., Chung, J., Kushman, N., Schrittwieser, J., Leblond, R., ... Vinyals, O. (2022). Competition-level code generation with AlphaCode. Science, 378(6624), 1092–1097.
6. Nijkamp, E., Pang, B., Hayashi, M., Tu, L., Wang, H., Zhou, Y., Savarese, S., & Xiong, C. (2023). CodeGen: An open large language model for code with multi-turn program synthesis. International Conference on Learning Representations.
7. Fried, D., Aghajanyan, A., Lin, J., Wang, S., Wallace, E., Shi, F., ... Lewis, M. (2023). InCoder: A generative model for code infilling and synthesis. International Conference on Learning Representations.
8. Yao, S., Zhao, J., Yu, D., Du, N., Shafran, I., Narasimhan, K., & Cao, Y. (2023). ReAct: Synergizing reasoning and acting in language models. International Conference on Learning Representations.
9. Schick, T., Dwivedi-Yu, J., Dessì, R., Raileanu, R., Lomeli, M., Zettlemoyer, L., Cancedda, N., & Scialom, T. (2023). Toolformer: Language models can teach themselves to use tools. Advances in Neural Information Processing Systems, 36.
10. Shinn, N., Cassano, F., Gopinath, A., Narasimhan, K., & Yao, S. (2023). Reflexion: Language agents with verbal reinforcement learning. Advances in Neural Information Processing Systems, 36.
11. Jiang, Hao, et al. "AgentOmnia: Scaling Agentic Models for Full-Scenario Applications." arXiv preprint arXiv:2607.23124 (2026).
12. Silver, D., Huang, A., Maddison, C. J., Guez, A., Sifre, L., van den Driessche, G., ... Hassabis, D. (2016). Mastering the game of Go with deep neural networks and tree search. Nature, 529(7587), 484–489.
13. Schulman, J., Wolski, F., Dhariwal, P., Radford, A., & Klimov, O. (2017). Proximal policy optimization algorithms. arXiv preprint arXiv:1707.06347.
14. Christiano, P. F., Leike, J., Brown, T., Martic, M., Legg, S., & Amodei, D. (2017). Deep reinforcement learning from human preferences. Advances in Neural Information Processing Systems, 30, 4299–4307.
15. Ouyang, L., Wu, J., Jiang, X., Almeida, D., Wainwright, C., Mishkin, P., ... Lowe, R. (2022). Training language models to follow instructions with human feedback. Advances in Neural Information Processing Systems, 35, 27730–27744.
16. Sculley, D., Holt, G., Golovin, D., Davydov, E., Phillips, T., Ebner, D., Chaudhary, V., Young, M., Crespo, J.-F., & Dennison, D. (2015). Hidden technical debt in machine learning systems. Advances in Neural Information Processing Systems, 28, 2503–2511.
17. Amershi, S., Begel, A., Bird, C., DeLine, R., Gall, H., Kamar, E., ... Zimmermann, T. (2019). Software engineering for machine learning: A case study. Proceedings of the 41st International Conference on Software Engineering: Software Engineering in Practice, 291–300.
18. Bender, E. M., Gebru, T., McMillan-Major, A., & Shmitchell, S. (2021). On the dangers of stochastic parrots: Can language models be too big? Proceedings of the 2021 ACM Conference on Fairness, Accountability, and Transparency, 610–623.
19. Weidinger, L., Uesato, J., Rauh, M., Griffin, C., Huang, P.-S., Mellor, J., ... Gabriel, I. (2022). Taxonomy of risks posed by language models. Proceedings of the 2022 ACM Conference on Fairness, Accountability, and Transparency, 214–229.
20. Strubell, E., Ganesh, A., & McCallum, A. (2019). Energy and policy considerations for deep learning in NLP. Proceedings of the 57th Annual Meeting of the Association for Computational Linguistics, 3645–3650.
Downloads
Published
Issue
Section
License
Copyright (c) 2026 Computer Science and Engineering Transactions
This work is licensed under a Creative Commons Attribution 4.0 International License.