Yinglun Xu

Pessimistic Reward Modeling in RLHF against Reward Hacking

SuperCoder: Assembly Program Superoptimization with Large Language Models

Robust Thompson Sampling Algorithms Against Reward Poisoning Attacks

Universal Black-Box Reward Poisoning Attack against Offline Reinforcement Learning

Binary Reward Labeling: Bridging Offline Preference and Reward-Based Reinforcement Learning

Two-Step Offline Preference-Based Reinforcement Learning with Constrained Actions