Improving Automated Code Reviews: Learning From Experience
Modern code review is a critical quality assurance process that is widely adopted in both industry and open source software environments. This process can help newcomers learn from the feedback of experienced reviewers; however, it often brings a large workload and stress to reviewers. To alleviate this burden, the field of automated code reviews aims to automate the process, teaching large language models to provide reviews on submitted code, just as a human would. A recent approach pre-trained and fine-tuned the code intelligent language model on a large-scale code review corpus. However, such techniques did not fully utilise quality reviews amongst the training data. Indeed, reviewers with a higher level of experience or familiarity with the code will likely provide deeper insights than the others. In this study, we set out to investigate whether higher-quality reviews can be generated from automated code review models that are trained based on an experience-aware oversampling technique. Through our quantitative and qualitative evaluation, we find that experience-aware oversampling can increase the correctness, level of information, and meaningfulness of reviews generated by the current state-of-the-art model without introducing new data. The results suggest that a vast amount of high-quality reviews are underutilised with current training strategies. This work sheds light on resource-efficient ways to boost automated code review models.
Mon 15 AprDisplayed time zone: Lisbon change
14:00 - 15:30 | Software QualityTechnical Papers / Registered Reports / Data and Tool Showcase Track at Grande Auditório Chair(s): Gopi Krishnan Rajbahadur Centre for Software Excellence, Huawei, Canada | ||
14:00 12mTalk | Not all Dockerfile Smells are the Same: An Empirical Evaluation of Hadolint Writing Practices by Experts Technical Papers Giovanni Rosa University of Molise, Simone Scalabrino University of Molise, Gregorio Robles Universidad Rey Juan Carlos, Rocco Oliveto University of Molise | ||
14:12 12mTalk | Supporting High-Level to Low-Level Requirements Coverage Reviewing with Large Language Models Technical Papers Anamaria-Roberta Hartl Johannes Kepler University Linz, Christoph Mayr-Dorn JOHANNES KEPLER UNIVERSITY LINZ, Atif Mashkoor Johannes Kepler University Linz, Alexander Egyed Johannes Kepler University Linz DOI Authorizer link Pre-print | ||
14:24 12mTalk | On the Executability of R Markdown Files Technical Papers Md Anaytul Islam Lakehead University, Muhammad Asaduzzman University of Windsor, Shaowei Wang Department of Computer Science, University of Manitoba, Canada | ||
14:36 12mTalk | APIstic: A Large Collection of OpenAPI Metrics Technical Papers souhaila serbout Software Institute @ USI, Cesare Pautasso Software Institute, Faculty of Informatics, USI Lugano | ||
14:48 6mTalk | Improving Automated Code Reviews: Learning From Experience Technical Papers Hong Yi Lin The University of Melbourne, Patanamon Thongtanunam University of Melbourne, Christoph Treude Singapore Management University, Wachiraphan (Ping) Charoenwet The University of Melbourne | ||
14:55 4mTalk | Multi-faceted Code Smell Detection at Scale using DesigniteJava 2.0 Data and Tool Showcase Track Tushar Sharma Dalhousie University Pre-print | ||
14:59 4mTalk | SATDAUG - A Balanced and Augmented Dataset for Detecting Self-Admitted Technical Debt Data and Tool Showcase Track Edi Sutoyo Bernoulli Institute for Mathematics, Computer Science and Artificial Intelligence, University of Groningen, Andrea Capiluppi University of Groningen | ||
15:03 4mTalk | Curated Email-Based Code Reviews Datasets Data and Tool Showcase Track Mingzhao Liang The University of Melbourne, Wachiraphan (Ping) Charoenwet The University of Melbourne, Patanamon Thongtanunam University of Melbourne | ||
15:07 4mTalk | TestDossier: A Dataset of Tested Values Automatically Extracted from Test Execution Data and Tool Showcase Track Andre Hora UFMG Pre-print Media Attached | ||
15:11 4mTalk | Greenlight: Highlighting TensorFlow APIs Energy Footprint Data and Tool Showcase Track Saurabhsingh Rajput Dalhousie University, Maria Kechagia University College London, Federica Sarro University College London, Tushar Sharma Dalhousie University Pre-print | ||
15:15 5mTalk | When Code Smells Meet ML: On the Lifecycle of ML-specific Code Smells in ML-enabled Systems Registered Reports Gilberto Recupito University of Salerno, Giammaria Giordano University of Salerno, Filomena Ferrucci University of Salerno, Dario Di Nucci University of Salerno, Fabio Palomba University of Salerno | ||
15:20 5mTalk | Comparison of Static Analysis Architecture Recovery Tools for Microservice Applications Registered Reports Simon Schneider Hamburg University of Technology, Alexander Bakhtin University of Oulu, Xiaozhou Li University of Oulu, Jacopo Soldani University of Pisa, Italy, Antonio Brogi Università di Pisa, Tomas Cerny University of Arizona, Riccardo Scandariato Hamburg University of Technology, Davide Taibi University of Oulu and Tampere University |