This graduate seminar is an in-depth reading group on vision-language models (VLMs), built around seminal papers. The arc starts from representational and pretraining fundamentals, moves through how vision gets connected to language models, then covers generation, grounding, retrieval, and agents, and closes on reasoning and evaluation. We meet twice a week over 12 weeks, with roughly one core paper per session and optional companions. Upon completion, students will have fluency across the full VLM stack, from contrastive pretraining to multimodal agents, along with the critical reading skills needed for research in the area.


1. Course overview

Prerequisites. There are no formal prerequisites, but this is not an introductory course. We read primary sources including transformer, diffusion, and discrete-representation papers directly, and discussion assumes you can follow the math and architecture without hand-holding. In practice, prior coursework or equivalent experience in machine learning and deep learning is a hard requirement. If you are unsure whether your background is sufficient, contact the instructor before enrolling.


2. Key dates

All deadlines are 11:59 p.m. ET on the date listed, submitted through LEARN.

Deliverable Due Weight
Weekly paper review Before each week's sessions, starting Week 2 20% (total)
In-class discussion Every session you attend 20%
Paper presentation Your assigned week (slides due 2 days prior) 20%
Project proposal Fri Oct 2, 2026 20%
Assignment 1 Fri Oct 23, 2026 10%
Project progress check-in Fri Nov 6, 2026 ungraded
Final report Wed Dec 8, 2026 20%
Assignment 2 Wed Dec 8, 2026 10%

Term calendar (Fall 2026). Classes begin Wednesday, Sep 9. Monday, Sep 7 is Labour Day, so the seminar's first meeting is Wednesday, Sep 9, which runs as a single instructor-led session. Reading week (no class) is Oct 10–18, and it absorbs Canadian Thanksgiving (Monday, Oct 12), so no Monday class meeting is lost to the holiday. Classes end Dec 8, and the final examination period runs Dec 10–22. Confirm against the University's official important-dates page; any changes will be announced in class and posted to the course website.


3. Grading

Component Weight
Paper reviews 20%
Paper presentation 20%
Post-presentation discussion 20%
Assignments (two × 10%) 20%
Course project 20%
Base total 100%
Bonus (see below) up to 10%
Maximum achievable 110%

Bonus (up to 10%, added on top of the base). There is exactly one bonus, split into two independent 5% halves: