Songtao Lu

Can Mamba Learn In Context with Outliers? A Theoretical Generalization Analysis

Aligning Frozen LLMs by Reinforcement Learning: An Iterative Reweight-then-Optimize Approach

Objective Soups: Multilingual Multi-Task Modeling for Speech Processing

Optimality and NP-Hardness of Transformers in Learning Markovian Dynamical Functions

Meta-D2AG: Causal Graph Learning with Interventional Dynamic Data

Training Nonlinear Transformers for Chain-of-Thought Inference: A Theoretical Generalization Analysis

Objective Soups: Multilingual Multi-Task Acoustic Modeling for Automatic Speech Recognition

TSP: A Two-Sided Smoothed Primal-Dual Method for Nonconvex Bilevel Optimization

DUET: Decentralized Bilevel Optimization without Lower-Level Strong Convexity

MMD-NSL: Mixed Multinomial Distribution-based Neuro-Symbolic Learning

Global Optimality of In-context Markovian Dynamics Learning