- author = "Alexander E. I. Brownlee and James Callan and Karine Even-Mendoza and Alina Geiger and Carol Hanna and Justyna Petke and Federica Sarro and Dominik Sobania",
- title = "Enhancing Genetic Improvement Mutations Using Large Language Models",
- booktitle = "SSBSE 2023: Challenge Track",
- year = "2023",
- editor = "Paolo Arcaini and Tao Yue and Erik Fredericks",
- organisers = "Erik Fredericks and Paolo Arcaini and Tao Yue and Rebecca Moussa and Thomas Vogel and Gregory Gay and Max Hort and Bobby R. Bruce and Jose Miguel Rojas and Vali Tawosi",
- volume = "14415",
- series = "LNCS",
- pages = "153--159",
- address = "San Francisco, USA",
- month = "8 " # dec,
- publisher = "Springer",
- keywords = "genetic algorithms, genetic programming, genetic improvement, GI, GIN, SBSE, LLM, AI, ANN, OpenAI, Langchain4J, Java, jCodec",
- isbn13 = "978-3-031-48795-8",
-
URL = "
https://arxiv.org/pdf/2310.19813.pdf",
-
URL = "
https://kclpure.kcl.ac.uk/portal/en/publications/enhancing-genetic-improvement-mutations-using-large-language-mode",
-
code_url = "https://doi.org/10.5281/zenodo.8304433",
-
code_url = "http://github.com/gintool/gin",
-
DOI = "
doi:10.1007/978-3-031-48796-5_13",
- size = "7 pages",
- abstract = "Large language models (LLMs) have been successfully applied to software engineering tasks, including program repair. However, their application in search-based techniques such as Genetic Improvement (GI) is still largely unexplored. we evaluate the use of LLMs as mutation operators for GI to improve the search process. We expand the Gin Java GI tool kit to call OpenAI API to generate edits for the JCodec tool. We randomly sample the space of edits using 5 different edit types. We find that the number of patches passing unit tests is up to 75percent higher with LLM-based edits than with standard Insert edits. Further, we observe that the patches found with LLMs are generally less diverse compared to standard edits. We ran GI with local search to find runtime improvements. Although many improving patches are found by LLM-enhanced GI, the best improving patch was found by standard GI",
-
notes = "Gin profiler
co-located with ESEC/FSE 2023. https://conf.researchr.org/track/ssbse-2023/ssbse-2023-challenge#Accepted-papers%gismo",
