Privately Aligning Language Models with Reinforcement Learning
This addresses privacy concerns in aligning LLMs for deployment, though it appears incremental by applying differential privacy to existing alignment methods.
The paper tackles the problem of aligning large language models with reinforcement learning while preserving privacy, introducing a differential privacy framework for both RL without human feedback and RLHF. Experimental results show the approach maintains competitive utility while providing strong privacy protections.
Positioned between pre-training and user deployment, aligning large language models (LLMs) through reinforcement learning (RL) has emerged as a prevailing strategy for training instruction following-models such as ChatGPT. In this work, we initiate the study of privacy-preserving alignment of LLMs through Differential Privacy (DP) in conjunction with RL. Following the influential work of Ziegler et al. (2020), we study two dominant paradigms: (i) alignment via RL without human in the loop (e.g., positive review generation) and (ii) alignment via RL from human feedback (RLHF) (e.g., summarization in a human-preferred way). We give a new DP framework to achieve alignment via RL, and prove its correctness. Our experimental results validate the effectiveness of our approach, offering competitive utility while ensuring strong privacy protections.