Beyond Similarity: A Gradient-based Graph Method for Instruction Tuning Data Selection

TL;DR

G2IS is a novel gradient-based graph method for instruction data selection that captures instruction interdependencies, improving domain adaptation efficiency and performance in instruction tuning of large language models.

Contribution

Introduces G2IS, a gradient-based graph approach that models instruction relationships for better data selection in domain-specific instruction tuning.

Findings

G2IS outperforms traditional data selection methods.

Significant performance gains in data-scarce domain adaptation tasks.

Enhanced training efficiency and effectiveness with G2IS.

Abstract

Large language models (LLMs) have shown great potential across various industries due to their remarkable ability to generalize through instruction tuning. However, the limited availability of domain-specific data significantly hampers their performance on specialized tasks. While existing methods primarily focus on selecting training data from general datasets that are similar to the target domain, they often fail to consider the joint distribution of instructions, resulting in inefficient learning and suboptimal knowledge transfer. To address these challenges, we introduce G2IS (Gradient-based Graph Instruction Selection), a novel method that constructs a mixed gradient-based instruction graph to capture the joint distribution and interdependencies between instructions. By accounting for the relationships between instructions, G2IS improves domain adaptation efficiency. Additionally, we propose a gradient walk algorithm to refine the data selection process, enhancing both training effectiveness and efficiency. Our experiments demonstrate that G2IS outperforms traditional methods across various domain adaptation tasks, yielding significant performance gains, particularly in complex, data-scarce scenarios. These results underscore the potential of G2IS in advancing the development of large, domain-specific models.