About The Seminar
The advances of foundation models appear increasingly omnipotent with the scaling of data and model size. However, their capabilities are inherently limited by their architectures. This seminar examines fundamental sequence modeling approaches from an architecture design perspective, focusing on issues such as efficiency, generalization, memorization, and expressiveness while enabling practical implementation on modern hardwares.
Sources
Want to present?
1. Get in touch
Contact us via email or any other method with the topic you'd like to present and your preferred time slot.
2. We add you
We will add you to an available time slot.
3. Present
Present your paper or a topic of interest.
Event Schedule
Organizers
This seminar is organized by

Songlin Yang
Massachusetts Institute of Technology

Malachy Yang
Carnegie Mellon University

Simran Arora
Stanford University