Saved in:
Bibliographic Details
Main Authors: Xu, Ruiyao, Samia, Noelle I., Liu, Han
Format: Preprint
Published: 2026
Subjects:
Online Access:https://arxiv.org/abs/2603.12932
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866911517535895552
author Xu, Ruiyao
Samia, Noelle I.
Liu, Han
author_facet Xu, Ruiyao
Samia, Noelle I.
Liu, Han
contents Adapting Large Language Models (LLMs) to specialized domains requires high-quality instruction tuning datasets, which are expensive to create through human annotation. Existing data synthesis methods focus on general-purpose tasks and fail to capture domain-specific terminology and reasoning patterns. To address this, we introduce DS$^2$-Instruct, a zero-shot framework that generates domain-specific instruction datasets without human supervision. Our approach first generates task-informed keywords to ensure comprehensive domain coverage. It then creates diverse instructions by pairing these keywords with different cognitive levels from Bloom's Taxonomy. Finally, it uses self-consistency validation to ensure data quality. We apply this framework to generate datasets across seven challenging domains, such as mathematics, finance, and logical reasoning. Comprehensive evaluation demonstrates that models fine-tuned on our generated data achieve substantial improvements over existing data generation methods.
format Preprint
id arxiv_https___arxiv_org_abs_2603_12932
institution arXiv
publishDate 2026
record_format arxiv
spellingShingle DS$^2$-Instruct: Domain-Specific Data Synthesis for Large Language Models Instruction Tuning
Xu, Ruiyao
Samia, Noelle I.
Liu, Han
Computation and Language
Adapting Large Language Models (LLMs) to specialized domains requires high-quality instruction tuning datasets, which are expensive to create through human annotation. Existing data synthesis methods focus on general-purpose tasks and fail to capture domain-specific terminology and reasoning patterns. To address this, we introduce DS$^2$-Instruct, a zero-shot framework that generates domain-specific instruction datasets without human supervision. Our approach first generates task-informed keywords to ensure comprehensive domain coverage. It then creates diverse instructions by pairing these keywords with different cognitive levels from Bloom's Taxonomy. Finally, it uses self-consistency validation to ensure data quality. We apply this framework to generate datasets across seven challenging domains, such as mathematics, finance, and logical reasoning. Comprehensive evaluation demonstrates that models fine-tuned on our generated data achieve substantial improvements over existing data generation methods.
title DS$^2$-Instruct: Domain-Specific Data Synthesis for Large Language Models Instruction Tuning
topic Computation and Language
url https://arxiv.org/abs/2603.12932