Pay More Attention to the Robustness of Prompt for Instruction Data Mining

2025-03-31

AI Summary

This study addresses the challenge of automatically selecting high-quality data for instruction tuning. We propose a novel data quality assessment paradigm centered on *prompt robustness*. Methodologically, we introduce the first online instruction data mining framework based on adversarial attacks: (1) generating semantically preserved but syntactically perturbed adversarial instructions; (2) modeling their *instruction-following difficulty*; and (3) measuring *embedding-space consistency* between model outputs for original and adversarial instructions—jointly forming our data filtering criterion. Experiments on two mainstream benchmarks demonstrate that our approach significantly improves downstream instruction-tuning performance (average +2.8% task accuracy), validating prompt robustness as an effective and practical proxy for data quality. The method provides a new, automated, and interpretable pathway for curating high-quality instruction data.

Technology Category

• Machine Learning: Adversarial Learning & Robustness
• Natural Language Processing: Safety and Robustness
• Computer Vision: Adversarial Attacks & Robustness

Application Category

• Economics, Online Markets and Human Computation: Data quality aspects of human-annotated datasets
• Web Mining and Content Analysis: Web data quality in the era of algorithmically-generated content
• Search and Retrieval-Augmented AI: Web evaluation methodologies and metrics

Abstract

Instruction tuning has emerged as a paramount method for tailoring the behaviors of LLMs. Recent work has unveiled the potential for LLMs to achieve high performance through fine-tuning with a limited quantity of high-quality instruction data. Building upon this approach, we further explore the impact of prompt's robustness on the selection of high-quality instruction data. This paper proposes a pioneering framework of high-quality online instruction data mining for instruction tuning, focusing on the impact of prompt's robustness on the data mining process. Our notable innovation, is to generate the adversarial instruction data by conducting the attack for the prompt of online instruction data. Then, we introduce an Adversarial Instruction-Following Difficulty metric to measure how much help the adversarial instruction data can provide to the generation of the corresponding response. Apart from it, we propose a novel Adversarial Instruction Output Embedding Consistency approach to select high-quality online instruction data. We conduct extensive experiments on two benchmark datasets to assess the performance. The experimental results serve to underscore the effectiveness of our proposed two methods. Moreover, the results underscore the critical practical significance of considering prompt's robustness.

Problem

• Explores prompt robustness impact on instruction data selection
• Proposes adversarial instruction data generation for quality mining
• Introduces metrics to evaluate adversarial data utility

Innovation

• Generates adversarial instruction data via prompt attacks
• Introduces Adversarial Instruction-Following Difficulty metric
• Proposes Adversarial Instruction Output Embedding Consistency method