Qiming Ge

large language model (2) continual learning (1) catastrophic forgetting (1) policy optimization (1) direct preference optimization (1) preference optimization (1) language model alignment (1) model safety (1) safety alignment (1) reinforcement learning from human feedback (1) human feedback (1) language model (1) low-rank adaptation (1) scaling law (1) downstream task (1) validation loss (1) refusal behavior (1) orthogonal subspace (1) overkill detection (1) contrastive decoding (1)