How does smaller and open-source Large Language Models compare to ChatGPT in refining code? Our study dives into code reviews, a cornerstone of modern software development. While code reviews are indispensable for ensuring quality and transferring knowledge, they can also become bottlenecks in large-scale projects.

• Inspired by a recent papaer by Guo et al, we explore how open-source models like CodeLlama and Llama 2 (7B parameters) measure up against proprietary solutions like ChatGPT for automating code refinement tasks.
• Our findings show that with proper tuning, these open-source models can offer an interesting balance between performance, cost-efficiency, and privacy.
• This research not only opens doors for privacy-conscious and cost-effective solutions but also sheds light on where current AI models shine—and where they still need a human touch.

Great news! Concordia has decided to fund the second edition of REANIMATE, our summer school on Retro Gaming History, Critic, and Development. This funds are part of the Aid to Research Related Events, Exhibition, Publication and Dissemination Activities (ARRE) Program.

The first edition of Reanimate'24 was organized by Prof. Yann Gael and team, and had a rich program, with 11 speakers from academia and industry, who shared their knowledge to participants of the event. The event included 5 full days of activities with talks, game jams, workshops, and it was a success. I am thrilled to join the organization for the second event, and glad that Concordia will be able to host the summer school again in 2025.

Yasmine presented her summer internship work at the Undergraduate Research Showcase at Concordia University. Her work entitled "Can ChatGPT Migrate My Code" explores the idea of using ChatGPT for migrating code that uses third-party libraries. The experiment consisted in prompting ChatGPT to migrate the code of one library version to another, and evaluating whether the generated code was correct. And the results were promising, with ChatGPT achieving a much higher degree of success than we originally anticipated. The poster presentation was a great success, way to go Yasmine! If you are interested in the details of this project, keep an eye out as we are preparing a paper submission soon.

As the internship period comes to an end, we gathered to bid a fond farewell to our interns, Adam and Yasmine. Over the past few months, their contributions have been invaluable, and they have truly become part of our lab family. To celebrate their hard work and dedication, we shared a delicious cake and exchanged heartfelt best wishes written on personalized cards. We are especially excited for Adam as he embarks on his upcoming internship at TMX, and we wish Yasmine tremendous success as she begins her new academic school year tomorrow. This isn't goodbye but more of a "see you later," as we hope to welcome both Adam and Yasmine back in the future. The day was filled with gratitude, memories, and well wishes for their bright futures ahead. We captured this special moment with a group picture in front of the cake, a token of appreciation for all they’ve brought to our team. Until we meet again, we wish Adam and Yasmine all the best in their future endeavors!

The efficiency of a Pull Request (PR) process hinges on how quickly maintainers and contributors respond to each other. Knowing how long this might take can improve interactions and manage expectations.

Our new study introduces a machine-learning method to predict these response times by analyzing data from 20 popular open-source GitHub projects. We examine various features of the projects and PRs, and identified key factors that influence response times.

• PRs submitted earlier in the week, with a moderate number of commits and clear descriptions, tend to get quicker responses.
• Contributors who are more engaged and have a good track record also tend to respond faster.
• We also highlight how understanding and predicting response times can enhance the PR review process.

The REALISE Lab had an amazing experience at the CREATE SE4AI retreat in Kingston this July 2024. The CREATE SE4AI is a training initiative focused on the development, deployment, and servicing of artificial intelligence-based software systems. The retreat provided us with a wonderful opportunity to connect with researchers and professors from the program, sharing our research ideas and experiences in person. We spent a day and night at the scenic Delta Inn hotel surrounded by beautiful landscapes. The retreat was highlighted by engaging presentations from students, including Rachna from our lab, who shared her work on her recently published paper at MSR 2024. The evening was capped off with a delightful dinner and a charming walk through the streets of Kingston, a town full of natural beauty and perfect for relaxation.

Following the retreat, we participated in the first day of the Canadian Software Engineering Research (CSER) conference. This event brought together renowned researchers in the field of software engineering. We gained new insights and perspectives from the talks, particularly those from young professors, which were both inspiring and informative. Prof. Costa's presentation on Dependency Management offered valuable insights into a critical area of our work. The conference provided us with valuable opportunities to network, learn, and collaborate with fellow researchers. The CREATE SE4AI retreat and CSER conference were memorable and enriching experiences for the REALISE Lab. They offered us the chance to connect with the broader research community, learn from leading experts, and showcase our work.

Finding performance regressions usually require the execution of long and costly performance test suites. This is because performance tests often have to test the system end-to-end. Could we reduce the testing costs by testing locally (e.g., module, a service, method) and using a model to predict the impact of local changes no the system as a whole?

Our new paper proposes exactly this! The paper entitled "Early Detection of Performance Regressions by Bridging Local Performance Data and Architectural Models" has been accepted at 47th IEEE/ACM International Conference on Software Engineering (ICSE 2025).

We are currently finalizing the camera-ready version of the paper, and we will share the preprint soon. Stay tuned for more updates!

Are you interested in training Chatbots for Software Engineering tasks? Our paper "A Transformer-based Approach for Augmenting Software Engineering Chatbots Datasets" has been accepted at 18th ACM/IEEE International Symposium on Empirical Software Engineering and Measurement (ESEM 2024).

• In this paper, we propose an approach to augment Chatbot training datasets tailored for Sofdrware Engineering tasks.

Zakaria has presented his work in progress entitled "Choosing the Right Bias Mitigation Strategy: Insights from a Comprehensive Benchmark." In this work, we conduct an experimental assessment of bias mitigation methods over 21 scenarios, considering different machine learning models and datasets.

Our goal is to set a large benchmarking framework to help practitioners choose the right bias mitigation strategy for their specific use case. The poster was presented at the 2024th edition of the Software Engineering for Machine Learning Applications (SEMLA).

I am thrilled to announce that FRQNT has funded our research program “Harnessing Software Ecosystems to Support Library Maintainers”, as part of the Support for New Academics program.

This is a 2-year program, designed to support maintainers in improving the reliability and quality of open- source software libraries by leveraging their dependent ecosystem. This project’s goal is to provide solutions that help maintainers

1. understand how dependents use their library to plan their library evolution and
2. harness the tests of the dependent ecosystem to improve the quality of their library project.