Saved in:
Bibliographic Details
Main Authors: Zhang, Lingyun, jin, Bin, Ge, Gaojian, Liu, Lunhui, Shen, Xuewen, Wu, Mingyong, Zhang, Houqian, Jiang, Yongneng, Chen, Shiqi, Pu, Shi
Format: Preprint
Published: 2024
Subjects:
Online Access:https://arxiv.org/abs/2406.11410
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866910492898885632
author Zhang, Lingyun
jin, Bin
Ge, Gaojian
Liu, Lunhui
Shen, Xuewen
Wu, Mingyong
Zhang, Houqian
Jiang, Yongneng
Chen, Shiqi
Pu, Shi
author_facet Zhang, Lingyun
jin, Bin
Ge, Gaojian
Liu, Lunhui
Shen, Xuewen
Wu, Mingyong
Zhang, Houqian
Jiang, Yongneng
Chen, Shiqi
Pu, Shi
contents Human priors play a crucial role in efficiently utilizing data in deep learning. However, with the development of large language models (LLMs), there is an increasing emphasis on scaling both model size and data volume, which often diminishes the importance of human priors in data construction. Influenced by these trends, existing Small Language Models (SLMs) mainly rely on web-scraped large-scale training data, neglecting the proper incorporation of human priors. This oversight limits the training efficiency of language models in resource-constrained settings. In this paper, we propose a principle to leverage human priors for data construction. This principle emphasizes achieving high-performance SLMs by training on a concise dataset that accommodates both semantic diversity and data quality consistency, while avoiding benchmark data leakage. Following this principle, we train an SLM named HARE-1.1B. Extensive experiments on large-scale benchmark datasets demonstrate that HARE-1.1B performs favorably against state-of-the-art SLMs, validating the effectiveness of the proposed principle. Additionally, this provides new insights into efficient language model training in resource-constrained environments from the view of human priors.
format Preprint
id arxiv_https___arxiv_org_abs_2406_11410
institution arXiv
publishDate 2024
record_format arxiv
spellingShingle HARE: HumAn pRiors, a key to small language model Efficiency
Zhang, Lingyun
jin, Bin
Ge, Gaojian
Liu, Lunhui
Shen, Xuewen
Wu, Mingyong
Zhang, Houqian
Jiang, Yongneng
Chen, Shiqi
Pu, Shi
Computation and Language
Artificial Intelligence
Human priors play a crucial role in efficiently utilizing data in deep learning. However, with the development of large language models (LLMs), there is an increasing emphasis on scaling both model size and data volume, which often diminishes the importance of human priors in data construction. Influenced by these trends, existing Small Language Models (SLMs) mainly rely on web-scraped large-scale training data, neglecting the proper incorporation of human priors. This oversight limits the training efficiency of language models in resource-constrained settings. In this paper, we propose a principle to leverage human priors for data construction. This principle emphasizes achieving high-performance SLMs by training on a concise dataset that accommodates both semantic diversity and data quality consistency, while avoiding benchmark data leakage. Following this principle, we train an SLM named HARE-1.1B. Extensive experiments on large-scale benchmark datasets demonstrate that HARE-1.1B performs favorably against state-of-the-art SLMs, validating the effectiveness of the proposed principle. Additionally, this provides new insights into efficient language model training in resource-constrained environments from the view of human priors.
title HARE: HumAn pRiors, a key to small language model Efficiency
topic Computation and Language
Artificial Intelligence
url https://arxiv.org/abs/2406.11410