Yao Shu

Assistant Professor@The Hong Kong University of Science and Technology (Guangzhou)·中国·OpenReview

研究方向

Large Language Models · Data-Centric AI · Optimization & Decision-Making · Learning Dynamic & Theory · Automated Machine Learning · Neural Architecture Search

Yao Shu

Model-based Offline RL via Robust Value-Aware Model Learning with Implicitly Differentiable Adaptive Weighting

FedPOB: Sample-Efficient Federated Prompt Optimization via Bandits

Test-Time Policy Adaptation for Enhanced Multi-Turn Interactions with LLMs

Scheduling Your LLM Reinforcement Learning with Reasoning Trees

1S-DAug: One-Shot Data Augmentation for Robust Few-Shot Generalization

Meta-Prompt Optimization for LLM-Based Sequential Decision Making

Adaptive Test-Time Compute Allocation via Training-Free Difficulty Proxies

Self-Reflective Generation at Test Time

T-POP: Test-Time Personalization with Online Preference Feedback

FURINA: A Fully Customizable Role-Playing Benchmark via Scalable Multi-Agent Collaboration Pipeline

Effective Policy Learning for Multi-Agent Online Coordination Beyond Submodular Objectives

ReDit: Reward Dithering for Improved LLM Policy Optimization

Refining Adaptive Zeroth-Order Optimization at Ease

Ferret: Federated Full-Parameter Tuning at Scale for Large Language Models

Multinoulli Extension: A Lossless Yet Effective Probabilistic Framework for Subset Selection over Partition Constraints

WMarkGPT: Watermarked Image Understanding via Multimodal Large Language Models

Ferret: Federated Full-Parameter Tuning at Scale for Large Language Models