TAG-INSTRUCT: Controlled Instruction Complexity Enhancement through Structure-based Augmentation

TL;DR

TAG-INSTRUCT introduces a structured, RL-guided method to enhance instruction complexity for large language models by compressing instructions into a tag space, enabling better control and stability.

Contribution

It proposes a novel tag-based framework for instruction complexity augmentation that outperforms previous prompt-based methods and improves controllability.

Findings

Outperforms existing instruction complexity augmentation methods

Operates in a compact tag space for better control

Provides stable and controllable instruction synthesis

Abstract

High-quality instruction data is crucial for developing large language models (LLMs), yet existing approaches struggle to effectively control instruction complexity. We present TAG-INSTRUCT, a novel framework that enhances instruction complexity through structured semantic compression and controlled difficulty augmentation. Unlike previous prompt-based methods operating on raw text, TAG-INSTRUCT compresses instructions into a compact tag space and systematically enhances complexity through RL-guided tag expansion. Through extensive experiments, we show that TAG-INSTRUCT outperforms existing instruction complexity augmentation approaches. Our analysis reveals that operating in tag space provides superior controllability and stability across different instruction synthesis frameworks.