【摘要】 [目的/意义]通过对国内外典型的大语言模型所生成的中文论文摘要进行对比分析，总结归纳两者间的异同点，为大语言模型后续的深度开发和发展研究提供参考。[方法/过程]选取2023年国家社会科学基金年度项目中“图书馆、情报与文献学”学科的121个课题作为题目，通过ChatGPT4.0与文心大模型4.0分别生成中文摘要，经过数据预处理及文本分析，从高频词特征、词性分布、句子数量以及摘要内容长度等角度探讨国内外大语言模型生成内容的异同。然后，与中文期刊《图书情报工作》中的摘要写作做对比，判断大语言模型生成摘要是否贴合中文论文写作规范。[结果/结论]文心一言生成摘要篇幅较短，字数较少，更贴合中文论文写作标准，GPT生成摘要的平均字数及句子数量较多，通过对比两个典型大语言模型生成内容的差距及特点，为大语言模型的完善与进一步深度开发提供一定的参考。更多还原

【Abstract】 [Purpose/Significance]By comparing and analyzing the abstracts of Chinese papers generated by typical Large Language Models at home and abroad,we summarize the similarities and differences between the two,and provide references for the subsequent in-depth development of the Large Language Models and the development of research.[Method/Process]121 topics in the discipline of “Library,Intelligence and Documentation “in the annual project of the National Social Science Foundation of China in 2023 were selected as the topics.The Chinese abstracts were generated by ChatGPT4.0 and ERNIE 4.0 respectively,and were analyzed in terms of the characteristics of high-frequency words,the distribution of words,the number of sentences,and the length of the abstract content to explore the similarities and differences of the content generated by the Large Language Models at home and abroad through the data preprocessing and the text analysis.Then,the comparison was also made with the abstracts written in the Chinese journal “Library and Intelligence Service” to determine whether the abstracts generated by the large language model are in line with the norms of Chinese thesis writing.[Result/Conclusion]The abstracts generated by ERNIE Bot are shorter,with fewer words,and more suitable for Chinese paper writing standards,while GPT generates abstracts with more words and sentences on average.By comparing the gaps and characteristics of the contents generated by the two typical Large Language Models,we can provide certain references for the improvement and further in-depth development of the Large Language Models.更多还原