Gespeichert in:
Bibliographische Detailangaben
Hauptverfasser: Shang, Zhaoyang, Wei, Sibo, Guo, Jianbin, Zhou, Rui, Dong, Lifeng, Luo, Yin
Format: Preprint
Veröffentlicht: 2025
Schlagworte:
Online-Zugang:https://arxiv.org/abs/2510.13892
Tags: Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
_version_ 1866914096002105344
author Shang, Zhaoyang
Wei, Sibo
Guo, Jianbin
Zhou, Rui
Dong, Lifeng
Luo, Yin
author_facet Shang, Zhaoyang
Wei, Sibo
Guo, Jianbin
Zhou, Rui
Dong, Lifeng
Luo, Yin
contents Large Language Models (LLMs) excel in general tasks, but adapting them to specialized domains relies on high-quality supervised fine-tuning (SFT) data. Although existing methods can identify subsets of high-quality data and reduce training cost to some extent, their selection process still suffers from over-reliance on LLMs' internal knowledge, weak interpretability, and limited generalization. To address these limitations, we propose THTB (The Harder The Better), a cognitive science-inspired framework for instruction data selection and annotation guidance. THTB prioritizes higher-level cognitive instructions by combining quality filtering with intrinsic and extrinsic hardness scoring, offering interpretable and quantifiable criteria for efficient SFT, both in data selection and annotation guidance. Experiments show that THTB enables models trained on only 5% of the data to outperform full-dataset training, while achieving superior generalization compared with LLM-only selection. In addition, THTB provides effective annotation guidance in vertical domains, enabling a model trained on just 2% of the data to surpass models trained on much larger datasets, demonstrating strong potential for domain adaptation. Our code, datasets, and models are available on https://github.com/DYJG-research/THTB.
format Preprint
id arxiv_https___arxiv_org_abs_2510_13892
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle The Harder The Better: Maintaining Supervised Fine-tuning Generalization with Less but Harder Data
Shang, Zhaoyang
Wei, Sibo
Guo, Jianbin
Zhou, Rui
Dong, Lifeng
Luo, Yin
Computation and Language
Large Language Models (LLMs) excel in general tasks, but adapting them to specialized domains relies on high-quality supervised fine-tuning (SFT) data. Although existing methods can identify subsets of high-quality data and reduce training cost to some extent, their selection process still suffers from over-reliance on LLMs' internal knowledge, weak interpretability, and limited generalization. To address these limitations, we propose THTB (The Harder The Better), a cognitive science-inspired framework for instruction data selection and annotation guidance. THTB prioritizes higher-level cognitive instructions by combining quality filtering with intrinsic and extrinsic hardness scoring, offering interpretable and quantifiable criteria for efficient SFT, both in data selection and annotation guidance. Experiments show that THTB enables models trained on only 5% of the data to outperform full-dataset training, while achieving superior generalization compared with LLM-only selection. In addition, THTB provides effective annotation guidance in vertical domains, enabling a model trained on just 2% of the data to surpass models trained on much larger datasets, demonstrating strong potential for domain adaptation. Our code, datasets, and models are available on https://github.com/DYJG-research/THTB.
title The Harder The Better: Maintaining Supervised Fine-tuning Generalization with Less but Harder Data
topic Computation and Language
url https://arxiv.org/abs/2510.13892