Automated Data Curation for Robust Language Model Fine-Tuning

TL;DR

This paper presents CLEAR, an automated data curation pipeline that improves fine-tuned language models by filtering and correcting training data based on confidence estimates, enhancing model performance without extra fine-tuning.

Contribution

The paper introduces CLEAR, a novel, data-centric framework for dataset curation that enhances LLM fine-tuning without requiring stronger models or additional training.

Findings

CLEAR improves model performance across multiple datasets and models

It effectively filters and corrects low-quality training data

The approach does not require access to stronger LLMs for data curation

Abstract

Large Language Models have become the de facto approach to sequence-to-sequence text generation tasks, but for specialized tasks/domains, a pretrained LLM lacks specific capabilities to produce accurate or well-formatted responses. Supervised fine-tuning specializes a LLM by training it on dataset of example prompts with target responses, but real-world data tends to be noisy. While many fine-tuning algorithms exist, here we consider a \emph{data-centric AI} perspective on LLM fine-tuning, studying how to \emph{systematically} curate the training dataset to improve the LLM produced via \emph{any} fine-tuning algorithm. We introduce an automated data curation pipeline CLEAR (Confidence-based LLM Evaluation And Rectification) for instruction tuning datasets, that can be used with any LLM and fine-tuning procedure. CLEAR estimates which training data is low-quality and either filters or corrects it. Automatically identifying which data to filter or correct is done via LLM-derived confidence estimates, to ensure only confident modifications to the dataset. Unlike existing data curation techniques, CLEAR is a comprehensive framework that can improve a dataset (and trained model outputs) without additional fine-tuning computations. We don't assume access to a stronger LLM than the model being fine-tuned (e.g.\ relying on GPT-4 when fine-tuning GPT-3.5), to see whether CLEAR can meaningfully improve the capabilities of any LLM. Experiments reveal that CLEAR consistently improves the performance of fine-tuned models across many datasets and models (like GPT-3.5 and Llama2).