ActiveUltraFeedback: Efficient Preference Data Generation using Active Learning

TL;DR

ActiveUltraFeedback introduces an active learning pipeline that efficiently generates high-quality preference data for training language models, reducing annotation costs while maintaining or improving performance.

Contribution

It presents a modular active learning framework with novel response selection methods, significantly reducing data annotation requirements for reinforcement learning from human feedback.

Findings

Achieves comparable or better performance with one-sixth of the data

Demonstrates effectiveness of novel response selection methods

Provides publicly available datasets and pipeline

Abstract

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.