@inproceedings{yuhang-etal-2024-yin,
    title = "银瞳:基于自适应语义空间学习的中文金融多任务大模型({S}ilver{S}ight: A Multi-Task {C}hinese Financial Large Language Model Based on Adaptive Semantic Space Learning)",
    author = "Zhou, Yuhang  and
      Li, Zeping  and
      Tian, Siyu  and
      Ni, Yuchen  and
      Zhang, Jian  and
      Liu, Xiang  and
      Ye, Guangnan  and
      Wu, Jie  and
      Chai, Hongfeng",
    editor = "Maosong, Sun  and
      Jiye, Liang  and
      Xianpei, Han  and
      Zhiyuan, Liu  and
      Yulan, He",
    booktitle = "Proceedings of the 23rd Chinese National Conference on Computational Linguistics (Volume 1: Main Conference)",
    month = jul,
    year = "2024",
    address = "Taiyuan, China",
    publisher = "Chinese Information Processing Society of China",
    url = "https://preview.aclanthology.org/author-degibert/2024.ccl-1.74/",
    pages = "955--972",
    language = "zho",
    abstract = "``大语言模型正逐渐被用于各种垂直领域,利用其广泛的知识储备来赋能领域中的多种场景。然而,各领域拥有多种待学习的特定任务,且多源异构的领域数据容易引发模型进行任务迁移时的冲突。基于此,本研究提出自适应语义空间学习框架,利用对语义空间内数据的自适应重分布,提升多专家模型的性能及选择效果,并基于此框架训练了一个金融多任务大模型{``}银瞳''。研究结果表明,我们的框架只需利用10{\%}的数据就能达到接近全数据训练的效果,并拥有较强的泛化表现。''"
}