ISoLA 2024

International Symposium On Leveraging Applications of Formal Methods, Verification and Validation • Crete, Greece

Talk

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

Time: Thursday, 31.10

Room: Room 3

Authors: Gerhard Stenzel, Kyrill Schmid, Michael Kölle, Philipp Altmann, Marian Lingsch-Rosenfeld, Maximilian Zorn, Tim Bücher, Thomas Gabor, Martin Wirsing, Lenz Belzner

Abstract: Current software development agents based on large language models (LLMs) are often defined using heuristic methods, which can limit their flexibility and e!ectiveness. 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.

Paper: SEGym_Optimizing_Large_Language_Model_Assisted_Software_Engineering_Agents_with_Reinforcement_Learning_-paper.pdf