Data Advisor: Dynamic Data Curation for Safety Alignment of Large Language Models

TL;DR

Data Advisor is a novel LLM-based framework that dynamically guides data generation to improve safety and coverage in large language model alignment, addressing quality and diversity issues in dataset creation.

Contribution

It introduces a data curation method that monitors and advises data generation based on predefined principles, enhancing safety and coverage in LLM datasets.

Findings

Improves safety alignment across multiple LLMs

Enhances data quality and diversity during dataset creation

Maintains model utility while increasing safety measures

Abstract

Data is a crucial element in large language model (LLM) alignment. Recent studies have explored using LLMs for efficient data collection. However, LLM-generated data often suffers from quality issues, with underrepresented or absent aspects and low-quality datapoints. To address these problems, we propose Data Advisor, an enhanced LLM-based method for generating data that takes into account the characteristics of the desired dataset. Starting from a set of pre-defined principles in hand, Data Advisor monitors the status of the generated data, identifies weaknesses in the current dataset, and advises the next iteration of data generation accordingly. Data Advisor can be easily integrated into existing data generation methods to enhance data quality and coverage. Experiments on safety alignment of three representative LLMs (i.e., Mistral, Llama2, and Falcon) demonstrate the effectiveness of Data Advisor in enhancing model safety against various fine-grained safety issues without sacrificing model utility.