A meeting of the SEMTL community will be held on Wednesday, April 1st, 2026 at 09:00. It will take place at Room 603, McConnell building, McGill University.
Registration
Please RSVP using this form.
Program
- 9:00-10:00: Keynote by Prof. Steven H.H. Ding (McGill): Beyond Embeddings: Learning to Represent and Reason About Software
Modern approaches to software analysis increasingly rely on learned representations. In binary code similarity detection, this evolution has progressed from handcrafted features to deep embeddings that map programs into high-dimensional vector spaces. While embeddings achieve strong performance, they remain opaque and difficult to interpret, limiting their usefulness in real-world analysis workflows.
In this talk, I argue that embeddings are not the end of the story. I first show how large language models can be adapted to produce robust vector representations of binary code, capturing semantic similarity across architectures, optimizations, and obfuscations. I then present a different paradigm: using language models to generate structured, human-interpretable features—such as input/output behavior, control flow patterns, and inferred functionality—that can be directly inspected, indexed, and reasoned about.
Finally, I demonstrate how these representations enable higher-level analysis, including reconstructing software lineage and tracking vulnerability propagation in complex supply chains. Together, these results suggest a shift from viewing representation as vectors to viewing it as a foundation for reasoning about software systems.
-
10:00-10:30: Coffee Break
-
10:30-12:40: Student Talks
| Name(s) | Paper | Institution | Links |
|---|---|---|---|
| Abdul Ali & Jazia Djoudad | Towards Self-Adaptive Languages: From Imprecise Patterns to Pattern Detection and Metamodel/Model Co-Evolution | McGill University | — |
| Bhagya Chembakottu | Lost in Security Tutorials? Finding Signals for Better Tutorial Selection | McGill University | Blog · Google Scholar |
| Billy Exarhakos | Choose Your Own Adventure: Non-Linear AI-Assisted Programming with EvoGraph | McGill University | Mila |
| Yifei (Byron) Chen | MIMIC: Integrating Diverse Personality Traits for Better Game Testing Using Large Language Model | McGill University | Paper · Website |
| Gagandeep Singh | Model-Driven Application Generation using Extended Domain Models | McGill University | Paper |
| Shuzhao Feng | Towards Repository-Level Code Generation with Structured Spec-Driven Engineering | McGill University | GitHub · LinkedIn |
| Sogol Masoumzadeh | SeBERTis: A Framework for Producing Classifiers of Security-Related Issue Reports | McGill University | Paper · LinkedIn |
| Zhuolin Xu | What Makes Code Generation Ethically Sourced? | Concordia University | Paper |
- 12:40: Social event (self-funded): 3 Brasseurs McGill. Address: 732 Rue Sainte-Catherine O, Montréal, QC H3B 4J8
Localisation
Room 603, McConnell Engineering Building, McGill University, 3480 Rue University, Montréal, QC H3A 2A7