Yingbin Liang

Provable In-Context Learning of Nonlinear Regression with Transformers

Mixture-of-Transformers Learn Faster: A Theoretical Study on Classification Problems

Discrete Diffusion Models: Novel Analysis and New Sampler Guarantees

Theory on Mixture-of-Experts in Continual Learning

Absorb and Converge: Provable Convergence Guarantee for Absorbing Discrete Diffusion Models

Multi-head Transformers Provably Learn Symbolic Multi-step Reasoning via Gradient Descent

Transformers Provably Learn Two-Mixture of Linear Classification via Gradient Flow

Theory on Score-Mismatched Diffusion Models and Zero-Shot Conditional Samplers

A Theoretical Analysis of Self-Supervised Learning for Vision Transformers

Broadening Target Distributions for Accelerated Diffusion Models via a Novel Analysis Approach

Dynamic Loss-Based Sample Reweighting for Improved Large Language Model Pretraining

How Transformers Learn Regular Language Recognition: A Theoretical Study on Training Dynamics and Implicit Bias

Model Developmental Safety: A Safety-Centric Method and Applications in Vision-Language Models

Unlocking the Power of Rehearsal in Continual Learning: A Theoretical Perspective

DUET: Decentralized Bilevel Optimization without Lower-Level Strong Convexity

Meta ControlNet: Enhancing Task Adaptation via Meta Learning

RLHF with Inconsistent Multi-Agent Feedback Under General Function Approximation: A Theoretical Perspective

Replay concurrently or sequentially? A theoretical perspective on replay in continual learning