Hangfan Zhang

large language model (2) adversarial learning (2) backdoor attack (2) reinforcement learning from human feedback (1) model alignment (1) poisoning attack (1) model ensemble (1) proximal policy optimization (1) reward shaping (1) jailbreak attack (1) graph contrastive learning (1) large language model alignment (1) llm alignment (1) group relative policy optimization (1) repeated sampling (1) graph neural network (1) trigger optimization (1) test-time compute (1) group preference reward shaping (1) preference-based reward (1)