ActiveUltraFeedback: Efficient Preference Data Generation using Active Learning
This addresses the efficiency bottleneck in aligning LLMs, particularly in low-resource and expert domains, by reducing annotation costs, though it is incremental as it builds on existing active learning and RLHF methods.
The paper tackles the high cost of acquiring preference data for RLHF by introducing ACTIVEULTRAFEEDBACK, an active learning pipeline that uses uncertainty estimates to select informative responses for annotation, achieving comparable or superior downstream performance with as little as one-sixth of the annotated data compared to static baselines.
Reinforcement Learning from Human Feedback (RLHF) has become the standard for aligning Large Language Models (LLMs), yet its efficacy is bottlenecked by the high cost of acquiring preference data, especially in low-resource and expert domains. To address this, we introduce ACTIVEULTRAFEEDBACK, a modular active learning pipeline that leverages uncertainty estimates to dynamically identify the most informative responses for annotation. Our pipeline facilitates the systematic evaluation of standard response selection methods alongside DOUBLE REVERSE THOMPSON SAMPLING (DRTS) and DELTAUCB, two novel methods prioritizing response pairs with large predicted quality gaps, leveraging recent results showing that such pairs provide good signals for fine-tuning. Our experiments demonstrate that ACTIVEULTRAFEEDBACK yields high-quality datasets that lead to significant improvements in downstream performance, notably achieving comparable or superior results with as little as one-sixth of the annotated data relative to static baselines. Our pipeline is available at https://github.com/lasgroup/ActiveUltraFeedback and our preference datasets at https://huggingface.co/ActiveUltraFeedback.