Theoretical exploration of foundation model adaption methods
Orchard View Room 330 N. Orchard Street, 3rd Floor NE, Madison, WisconsinSpeaker: Kangwook Lee
Loading view.
IFDS is located in the Central and Pacific Time zone. Please note the zone when accessing a particular event.
SILO: Theoretical Exploration of Foundation Model Adaptation Methods
Orchard View Room 330 N. Orchard Street, 3rd Floor NE, Madison, WisconsinKangwook Lee
Offline Multi-task Transfer RL with Representational Penalization
CSE (Allen) 403Speaker Bio: Avinandan is a second year PhD student, advised >
A good score does not lead to a good generative model
WID 1145 330 N Orchard Street, Madison, WISpeaker: Sixu Li Abstract: Score-based Generative Models (SGMs) is one leading >
SILO: Foundations of Real-World Reinforcement Learning
Orchard View Room 330 N. Orchard Street, 3rd Floor NE, Madison, WisconsinJeongyeol Kwon
GumbelSpec Sampling for Accelerating LLM Inference
CSE (Allen) 403Bio: Tianxiao Shen is a postdoctoral scholar at the University of Washington, >
Prelimit coupling and steady-state convergence of constant-stepsize nonsmooth contractive SA
WID 1145 330 N Orchard Street, Madison, WISpeaker: Yixuan Zhang Abstract: Motivated by Q-learning, we study nonsmooth >
SILO: Reinforcement Learning with Robustness and Safety Guarantees
Orchard View Room 330 N. Orchard Street, 3rd Floor NE, Madison, WisconsinDileep Kalathil, TAMU
IFDS Ideas Forum Double Header
Orchard View Room 330 N. Orchard Street, 3rd Floor NE, Madison, WisconsinSpeaker 1: Puqian Wang Robustly Learning Single-Index Models via Alignment >
SILO: Online Learning Guided Quasi-Newton Methods: Improved Global Non-asymptotic Guarantees
Orchard View Room 330 N. Orchard Street, 3rd Floor NE, Madison, WisconsinAryan Mokhtari, UT
Low-Rank Structures in Optimal Transport
CSE (Allen) 403Bio: Meyer Scetbon is currently a Research Scientist at Microsoft Research. >
Towards Plurality: Learning from Diverse Human Preferences
Orchard View Room 330 N. Orchard Street, 3rd Floor NE, Madison, WisconsinSpeaker: Ramya Vinayak Abstract: Large pre-trained models trained on internet-scale >