Self-Play with Adversarial Critic: Provable and Scalable Offline Alignment for Language Models
원문 초록
This work studies the challenge of aligning large language models (LLMs) with offline preference data. We focus on alignment by Reinforcement Learning from Human Feedback (RLHF) in particular. While popular preference optimization methods exhibit good empirical performance in practice, they are not theoretically guaranteed to converge to the optimal policy and can provably fail when the data coverage is sparse by classical offline reinforcement learning (RL) results. On the other hand, a recent line of work has focused on theoretically motivated preference optimization methods with provable guarantees, but these are not computationally efficient for large-scale applications like LLM alignment. To bridge this gap, we propose SPAC, a new offline preference optimization method with self-play, inspired by the on-average pessimism technique from the offline RL literature, to be the first provable and scalable approach to LLM alignment. We both provide theoretical analysis for its convergence under single-policy concentrability for the general function approximation setting and demonstrate its competitive empirical performance for LLM alignment on a 7B Mistral model with Open LLM Leaderboard evaluations.
함께 살펴볼 논문
공통 태그를 근거로 연결했습니다. 실제 인용 관계나 논문의 우수성을 뜻하지 않습니다.
Margin Trader: A Reinforcement Learning Framework for Portfolio Management with Margin and Constraints
Task Independent Capsule-Based Agents for Deep Q-Learning
Accelerating lifelong reinforcement learning via reshaping rewards
Federated Model Search via Reinforcement Learning
Variation-resistant Q-learning: Controlling and Utilizing Estimation Bias in Reinforcement Learning for Better Performance
A State Representation Dueling Network for Deep Reinforcement Learning
토론
이 논문과 연결된 포럼 글이 아직 없습니다.