Xuehai Pan

reinforcement learning from human feedback (3) safe reinforcement learning (2) large language model (2) preference learning (2) constraint optimization (1) knowledge distillation (1) content moderation (1) policy learning (1) language model alignment (1) ai safety (1) safety alignment (1) risk minimization (1) constraint satisfaction (1) human feedback (1) bayesian network (1) hallucination reduction (1) safety benchmark (1) alignment method (1) model-agnostic approach (1) residual correction (1)