On the Automation of Code Review Tasks Through Cross-Task Knowledge Distillation
Thu 1 May 2025 11:48 - 11:54 at 204 - ACM Student Research Presentations Chair(s): Md Tajmilur Rahman, Lola Burgueño
Code review is essential for ensuring code quality and minimizing errors, but it is often complex, subjective, and time-consuming. Existing research treated code review tasks (\ie \emph{quality estimation}, \emph{comment generation}, and \emph{code refinement}) separately, despite their interconnections. We propose a novel deep learning architecture using cross-task knowledge distillation to address these tasks simultaneously. Our approach outperforms state-of-the-art methods in \emph{comment generation} and \emph{code refinement}, achieving superior BLEU and CodeBLEU scores.
Tue 29 AprDisplayed time zone: Eastern Time (US & Canada) change
Thu 1 MayDisplayed time zone: Eastern Time (US & Canada) change
11:00 - 12:30 | ACM Student Research PresentationsSRC - ACM Student Research Competition at 204 Chair(s): Md Tajmilur Rahman , Lola Burgueño University of Malaga A subset of finalist ACM SRC students will give short presentations in this session. That decision about who will present will be made after the poster sessions, and this schedule will be updated, so don’t rely on the precise timing until just before the session.. They all also have posters in Canada Hall 3 Poster area, with judging to be on Tuesday. Awards will be announced in the banquet on Thursday evening. | ||
11:00 6mTalk | Automatic Fuzz Drivers for JavaScript with Type Distributions SRC - ACM Student Research Competition Mayant Mukul University of British Columbia | ||
11:06 6mTalk | CASS: Context-Aware Slice Summarization for Debugging Regression Failures SRC - ACM Student Research Competition Sahar Badihi University of British Columbia, Canada Pre-print | ||
11:12 6mTalk | Characterising Algorithm Debt in Machine and Deep Learning Systems SRC - ACM Student Research Competition Emmanuel Iko-Ojo Simon Australian National University | ||
11:18 6mTalk | Consistent Graph Model Generation with Large Language Models SRC - ACM Student Research Competition Boqi Chen McGill University | ||
11:24 6mTalk | Enhancing OSS Remediation with Patch Backporting SRC - ACM Student Research Competition Lyuye Zhang Nanyang Technological University | ||
11:30 6mTalk | Identifying Performance-Sensitive Configurations in Software Systems with LLM-Driven Agents SRC - ACM Student Research Competition Zehao Wang Concordia University | ||
11:36 6mTalk | Improving Formal Methods VisualizationsFormal Methods SRC - ACM Student Research Competition Avinash Palliyil Georgia Institute of Technology | ||
11:42 6mTalk | MUARF: Leveraging Multi-Agent Workflows for Automated Code Refactoring SRC - ACM Student Research Competition Yisen Xu Software PErformance, Analysis, and Reliability (SPEAR) lab, Concordia University, Montreal, Canada | ||
11:48 6mTalk | On the Automation of Code Review Tasks Through Cross-Task Knowledge Distillation SRC - ACM Student Research Competition Oussama Ben Sghaier DIRO, Université de Montréal | ||
11:54 6mTalk | On the Fly Input Refinement for Code Language Models SRC - ACM Student Research Competition Ravishka Shemal Rathnasuriya University of Texas at Dallas | ||
12:00 6mTalk | Program Feature-based Fuzzing Benchmarking SRC - ACM Student Research Competition Miao Miao The University of Texas at Dallas | ||
12:06 6mTalk | Revisiting SWE-Bench: On the Importance of Data Quality for LLM-based Code Models SRC - ACM Student Research Competition Reem Aleithan York University, Canada | ||
12:12 6mTalk | The Balancing Act of Policies in Developing Machine Learning Explanations SRC - ACM Student Research Competition Jacob Tjaden Colby College | ||
12:18 6mTalk | To Mock or Not to Mock: Divergence in Mocking Practices Between LLM and Developers SRC - ACM Student Research Competition Hanbin Qin Stevens Institute of Technology | ||
12:24 6mTalk | Towards Compatibly Mitigating Technical Lag in Maven Projects SRC - ACM Student Research Competition Rui Lu East China Normal University |