Reinforcement Learning from Human Feedback with Active Queries
This addresses the expensive data collection problem in RLHF for LLM alignment, offering a practical improvement for AI developers.
The paper tackles the high cost of collecting human preference data for aligning large language models by proposing query-efficient reinforcement learning from human feedback methods, achieving performance matching state-of-the-art DPO with about half the queries.
Aligning large language models (LLM) with human preference plays a key role in building modern generative models and can be achieved by reinforcement learning from human feedback (RLHF). Despite their superior performance, current RLHF approaches often require a large amount of human-labelled preference data, which is expensive to collect. In this paper, inspired by the success of active learning, we address this problem by proposing query-efficient RLHF methods. We first formalize the alignment problem as a contextual dueling bandit problem and design an active-query-based proximal policy optimization (APPO) algorithm with an $\tilde{O}(d^2/Δ)$ instance-dependent regret bound and an $\tilde{O}(d^2/Δ^2)$ query complexity, where $d$ is the dimension of feature space and $Δ$ is the sub-optimality gap over all the contexts. We then propose ADPO, a practical version of our algorithm based on direct preference optimization (DPO) and apply it to fine-tuning LLMs. Our experiments show that ADPO, while only making about half of queries for human preference, matches the performance of the state-of-the-art DPO method.