SEGym: Optimizing Large Language Model Assisted Software Engineering Agents with Reinforcement Learning

Stenzel, Gerhard; Schmid, Kyrill; Kölle, Michael; Altmann, Philipp ; Lingsch-Rosenfeld, Marian ; Zorn, Maximilian ; Bücher, Tim; Gabor, Thomas ; Wirsing, Martin und Belzner, Lenz (2025): SEGym: Optimizing Large Language Model Assisted Software Engineering Agents with Reinforcement Learning. AISoLA 2024: Second International Conference, Chersonissos, Griechenland, 30. Oktober - 03. November 2024. Bernhard, Steffen (Hrsg.): In: Bridging the Gap Between AI and Reality : Second International Conference, AISoLA 2024, Crete, Greece, October 30 – November 3, 2024, Proceedings, Cham: Springer. S. 107-124

Abstract

Current software development agents based on large language models (LLMs) are often defined using heuristic methods, which can limit their flexibility and effectiveness. Moreover, the entry barriers for new researchers in this field are high, largely due to the complex infrastructure required to develop and optimize these agents. This paper proposes a new approach: modeling software development agents over LLMs as a partially observable Markov decision process (POMDP) to enable data-driven optimization. To support this approach, we introduce SEGym, a framework based on the Gym interface for reinforcement learning agents. SEGym simplifies the setup of optimization experiments for software development agents within the POMDP framework, making it more accessible for researchers to engage in this field.

• Dokumententyp: Konferenzbeitrag (Paper)
• Fakultät: Mathematik, Informatik und Statistik > Informatik
• Themengebiete: 000 Informatik, Informationswissenschaft, allgemeine Werke > 004 Informatik
• ISBN: 978-3-031-75434-0
• Ort: Cham