Zhuoran Yang

Dual-Robust Cross-Domain Offline Reinforcement Learning Against Dynamics Shifts

Muon Outperforms Adam in Tail-End Associative Memory Learning

Cross-domain Offline Policy Adaptation with Dynamics- and Value-aligned Data Filtering

Taming Polysemanticity in LLMs: Theory-Grounded Feature Recovery via Sparse Autoencoders

Interpreting Multi-Layer Transformers for In-Context Linear Regression with Varying Covariance

How Transformers Learn Causal Structures In-Context: Explainable Mechanism Meets Theoretical Guarantee

Unlocking Out-of-Distribution Generalization in Transformers via Latent Space Reasoning

AV-Odyssey Bench: From Fundamental Audio Perception to Audio-Visual Understanding

Learning to Incentivize on the Fly: Leader-Follower Games with Policy Recommendation

Learning in Context, Guided by Choice: A Reward-Free Paradigm for Reinforcement Learning with Transformers

On the Mechanism and Dynamics of Modular Addition: Fourier Features, Lottery Ticket, and Grokking

Mechanistic Interpretability of In-Context Learning Generalization through Structured Task Curriculum

Can VLMs Reason Through Multiple Views?

Can Neural Networks Achieve Optimal Computational-statistical Tradeoff? An Analysis on Single-Index Model

Decoding Rewards in Competitive Games: Inverse Game Theory with Entropy Regularization

In-Context Linear Regression Demystified: Training Dynamics and Mechanistic Interpretability of Multi-Head Softmax Attention

Provable Learning for DEC-POMDPs: Factored Models and Memoryless Agents

In-Context Reinforcement Learning From Suboptimal Historical Data

In-Context Reinforcement Learning From Suboptimal Historical Data

Exploration in the Face of Strategic Responses: Provable Learning of Online Stackelberg Games

The Sample Complexity of Online Strategic Decision Making with Information Asymmetry and Knowledge Transportability

STRIDE: A Tool-Assisted LLM Agent Framework for Strategic and Interactive Decision-Making

BanditSpec: Adaptive Speculative Decoding via Bandit Algorithms

Quantile-Optimal Policy Learning under Unmeasured Confounding

An Instrumental Value for Data Production and its Application to Data Pricing

Steer a Crowd: Learning to Persuade a Population in a Stackelberg Game