2023/06/04 by Banghua Zhu, Zhu, Banghua, Hiteshi Sharma +11 · 2 citations
Computer Science · #Artificial Intelligence (cs.AI) #Computation and Language (cs.CL) #FOS: Computer and information sciences #FOS: Electrical engineering #Machine Learning (cs.LG) #Reinforcement Learning in Robotics #Systems and Control (eess.SY) #Topic Modeling #electronic engineering #information engineering
paper · pdf · doi:10.48550/arxiv.2306.02231
openalex publication_date 2023/06/04 · openalex created_date 2025/10/10 · openalex updated_date 2026/07/28
Reinforcement learning from human feedback (RLHF) has emerged as a reliable approach to aligning large language models (LLMs) to human preferences. Among the plethora of RLHF techniques, proximal policy optimization (PPO) is of the most widely used methods. Despite its popularity, however, PPO may suffer from mode collapse, instability, and poor sample efficiency. We show that these issues can be alleviated by a novel algorithm that we refer to as Advantage-Induced Policy Alignment (APA), which leverages a squared error loss function based on the estimated advantages. We demonstrate empirically that APA consistently outperforms PPO in language tasks by a large margin, when a separate reward model is employed as the evaluator. In addition, compared with PPO, APA offers a more stable form of control over the deviation from the model's initial policy, ensuring that the model improves its performance without collapsing to deterministic output. In addition to empirical results, we also provide a theoretical justification supporting the design of our loss function.