[OOPSLA24] PyDex: Repairing Bugs in Introductory Python Assignments using LLMs скачать в хорошем качестве

[OOPSLA24] PyDex: Repairing Bugs in Introductory Python Assignments using LLMs

PyDex: Repairing Bugs in Introductory Python Assignments using LLMs (Video, OOPSLA 2024) Jialu Zhang, José Pablo Cambronero, Sumit Gulwani, Vu Le, Ruzica Piskac, Gustavo Soares, and Gust Verbruggen (University of Waterloo, Canada; Microsoft, USA; Microsoft, USA; Microsoft, USA; Yale University, USA; Microsoft, USA; Microsoft, Belgium) Abstract: Students often make mistakes in their introductory programming assignments as part of their learning process. Unfortunately, providing custom repairs for these mistakes can require a substantial amount of time and effort from class instructors. Automated program repair (APR) techniques can be used to synthesize such fixes. Prior work has explored the use of symbolic and neural techniques for APR in the education domain. Both types of approaches require either substantial engineering efforts or large amounts of data and training. We propose to use a large language model trained on code, such as Codex (a version of GPT), to build an APR system – PyDex – for introductory Python programming assignments. Our system can fix both syntactic and semantic mistakes by combining multi-modal prompts, iterative querying, test-case-based selection of few-shots, and program chunking. We evaluate PyDex on 286 real student programs and compare to three baselines, including one that combines a state-of-the-art Python syntax repair engine, BIFI, and a state-of-the-art Python semantic repair engine for student assignments, Refactory. We find that PyDex can fix more programs and produce smaller patches on average. Article: https://doi.org/10.1145/3649850 ORCID: https://orcid.org/0009-0003-8193-0719, https://orcid.org/0000-0002-0713-6141, https://orcid.org/0000-0002-9226-9634, https://orcid.org/0000-0003-3727-3291, https://orcid.org/0000-0002-3267-0776, https://orcid.org/0000-0002-8061-9000, https://orcid.org/0000-0001-9182-597X Video Tags: AI for programming education, large language models, automated program repair, oopslaa24main-p144-p, doi:10.1145/3649850, orcid:0009-0003-8193-0719, orcid:0000-0002-0713-6141, orcid:0000-0002-9226-9634, orcid:0000-0003-3727-3291, orcid:0000-0002-3267-0776, orcid:0000-0002-8061-9000, orcid:0000-0001-9182-597X Presentation at the OOPSLA 2024 conference, October 20–25, 2024, https://2024.splashcon.org/track/spla... Sponsored by ACM SIGPLAN,