Consortium for Software Engineering Research
2026 Spring Meeting

Keynotes

Song Wang

York University

Understanding and Detecting AI Infrastructure Bugs

Talk Abstract: AI infrastructure software, such as TensorFlow and PyTorch, forms the foundation of modern machine learning systems. However, the scale, complexity, and rapidly evolving nature of these frameworks introduce significant challenges in ensuring their reliability, robustness, and security. In this talk, I will present our recent efforts on understanding, detecting, and mitigating bugs and vulnerabilities in AI infrastructure software. I will begin by discussing empirical studies that characterize the unique properties and root causes of bugs and vulnerabilities in AI infrastructure systems. I will then examine the limitations of existing bug and vulnerability detection techniques, including static analysis and dynamic fuzzing, and highlight why many traditional approaches struggle to effectively detect AI-specific bugs. Next, I will introduce our efforts for automatically uncovering bugs and vulnerabilities in deep learning libraries. Finally, I will discuss our pilot studies on demystifying and detecting bugs in AI agents, along with the emerging challenges and opportunities in building reliable and trustworthy agentic AI systems.

Bio: Song Wang is an Associate Professor at York University, Canada. He currently serves as an Associate Editor for ACM Transactions on Software Engineering and Methodology (TOSEM) and Empirical Software Engineering (EMSE). His research focuses on building reliable AI infrastructure. His work has uncovered hundreds of bugs in widely used AI frameworks such as TensorFlow and PyTorch. He has also proposed behavior-driven testing techniques for AI agents, which have identified more than 600 defects in popular AI coding tools, including Claude Code 4.6, Codex, and Gemini CLI. His work has received four best paper awards, including a Distinguished Paper Award at APSEC 2023, an ACM Distinguished Paper Award at ICPC 2022, an ACM Distinguished Paper Award at ICSE 2020, and a Best Paper Award at PROMISE 2019. He also received the ICSE Most Influential Paper Award in 2026. He is the recipient of the 2025 CS-Can | Info-Can Outstanding Early Career Computer Science Researcher Prize.

Website: Personal Website

Dayi Lin

Microsoft

Beyond the Core Loop: Variants and Invariants in Coding Agents

Talk Abstract: As foundation models grow stronger, coding agent architectures churn rapidly: workflows, tools, and orchestration frameworks that were state-of-the-art yesterday are being absorbed into model weights today. But beneath this architectural volatility, certain challenges prove stubbornly persistent. In this talk, I argue that the most valuable design investments for coding agents lie not in optimizing any particular scaffold, but in understanding the invariants: the structural properties that remain true regardless of how capable models become. I present a set of research that explores what will change from what will not, and what that means for building coding agents that are reliable, debuggable, and useful.

Bio: Dayi Lin is a Principal Applied Scientist at Microsoft. His research is at the intersection of software engineering and AI, including trustworthy coding agents and improving the software engineering capabilities of LLMs. He holds a PhD in Software Engineering from Queen's University, Canada.

CSER 2026 (Spring) Program

Note: The program below is tentative and subject to change. Session topics and paper titles will be finalized closer to the event.

Invited New Faculty Talks

Details of invited new faculty talks will be announced soon.

Panel

Panel details will be announced soon.

Tutorial

Tutorial details will be announced soon.

Research Talks

Research talk details will be announced soon.

Posters/Demos

Poster and demo details will be announced soon.

Dear CSER Community,

We are excited to announce that CSER 2026 Spring is on the horizon! We are soliciting proposals for short research talks. Speakers will introduce themselves, share their research, and engage with the community. Each talk will last 20 minutes, followed by a brief Q&A session. Please note that all talks will be held in-person at the University of Alberta.

This is an excellent opportunity to present your work to a broader audience and foster academic collaboration.

Faculty members, postdoctoral researchers, and industry researchers may all submit using the same link below. If you are interested in speaking, please include:

• Your name
• Your institution
• Email contact information
• A tentative title for your talk

Please complete the Faculty Talk Proposal Form by Tuesday, May 12, 2026. After the submission deadline, we will follow up with further details regarding the schedule and presentation format.

If you know colleagues who might be interested, please forward this invitation to them or contact us directly.

We look forward to your participation!

Call for Student Presentation/Poster

Dear CSER Community,

We also invite students to showcase their research at CSER 2026 Spring. You may choose to deliver a 10–15 minute presentation or submit a poster/demo. This is a fantastic opportunity to gain presentation experience and connect with the software engineering research community.

For your submission, please include:

• Your name
• Name of your research supervisor
• Institution, Program, Year of Study
• Email contact information
• Title of the talk/poster
• A 250 words abstract for the talk/poster

Please submit your proposal using the Student Presentation/Poster Proposal Form by Tuesday, May 12, 2026. We will provide the final schedule and additional presentation details after the submission deadline.

Faculty members, please encourage your students to participate!

We look forward to your innovative submissions!

Organizing Committee

Program Co-Chairs

• Lei Ma, University of Alberta
• Zhou Yang, University of Alberta

Web & Publicity Chair

• Zixi Chai, University of Alberta
• Jiaxuan Tong, University of Alberta

CSER Steering Committee

• Marin Litoiu, York
• Kelly Lyons, Toronto
• Hausi Müller, Victoria

Venue

University of Alberta

• University Commons,
  University of Alberta, Edmonton, AB, Canada

Registration

Registration and fees will be handled through Eventbrite. Planned rates are shown below. You can follow the Registration Link to register your ticket.

How to register:

Registration is completed on Groupize. A typical flow is:

1. Open the Registration Link.
2. Enter your email address, first name, and last name in the form on that site.
3. Click Register to submit. If you have already started, use 'Already Registered?' on the same page to request another sign-in email.
4. Check your inbox for an email with an access link. Open the link, then follow the on-screen steps to finish or update your registration.
5. On the event site, select your tickets and confirm your registration.
6. Proceed to checkout and complete payment.

Cancellation Policy:

• A $100 cancellation fee applies.
• Last date for cancellation: May 17, 2026.
• Registration transfer to another participant (same category) is allowed until May 17, 2026.

Supporters