Article(id=1329460968599433730, tenantId=1146029695717560320, journalId=1324759914146529478, issueId=1329460877864055544, articleNumber=null, orderNo=null, doi=10.7544/issn1000-1239.202330801, pmid=null, cstr=null, oa=null, hot=null, price=null, onlineType=0, articleFormat=0, articleType=null, articleTypeStr=null, receivedDate=1696867200000, receivedDateStr=2023-10-10, revisedDate=1710172800000, revisedDateStr=2024-03-12, acceptedDate=null, acceptedDateStr=null, onlineDate=1794897795104, onlineDateStr=2026-11-17, pubDate=1714492800000, pubDateStr=2024-05-01, doiRegisterDate=null, doiRegisterDateStr=null, onlineIssueDate=1794897795104, onlineIssueDateStr=2026-11-17, onlineJustAcceptDate=null, onlineJustAcceptDateStr=null, onlineFirstDate=null, onlineFirstDateStr=null, sourceXml=null, magXml=null, createTime=1794897795104, creator=13701087609, updateTime=1794897795104, updator=13701087609, issue=Issue{id=1329460877864055544, tenantId=1146029695717560320, journalId=1324759914146529478, year='2024', volume='61', issue='5', pageStart='1085', pageEnd='1347', issueExtLink='null', onlineDate='null', pubDate='1714492800000', pubDateStr='2024-05-01', beforeIssueId=null, nextIssueId=null, price=null, status=1, issueComplete=1, articleOrder=1, issueType=-1, specialIssue=null, createTime=1794897773471, creator='13701087609', updateTime=1794897773471, updator='13701087609', preIssue=null, nextIssue=null, articleTotal=null, ext=null, issueFiles=null, downloadFileDto=null}, startPage=1128, endPage=1142, ext={EN=ArticleExt(id=1329460968767205891, articleId=1329460968599433730, tenantId=1146029695717560320, journalId=1324759914146529478, language=EN, title=Robustness of GPT Large Language Models on Natural Language Processing Tasks, columnId=1329460878547727097, journalTitle=Journal of Computer Research and Development, columnName=Special Issue on Large Models and Safety, runingTitle=null, highlight=null, articleAbstract=

The GPT models have demonstrated impressive performance in various natural language processing (NLP) tasks. However, their robustness and abilities to handle various complexities of the open world have not yet to be well explored, which is especially crucial in assessing the stability of models and is a key aspect of trustworthy AI. In this study, we perform a comprehensive experimental analysis of GPT-3 and GPT-3.5 series models, exploring their performance and robustness using 15 datasets (about 147000 original test samples) with 61 robust probing transformations from TextFlint covering 9 popular NLP tasks. Additionally, we analyze the model’s robustness across different transformation levels, including character, word, and sentence. Our findings reveal that while GPT models exhibit competitive performance in certain tasks like sentiment analysis, semantic matching, and reading comprehension, they exhibit severe confusion regarding information extraction tasks. For instance, GPT models exhibit severe confusion in relation extraction and even exhibit “hallucination” phenomena. Moreover, they experience significant degradation in robustness in terms of tasks and transformations, especially in classification tasks and sentence-level transformations. Furthermore, we validate the impact of the quantity and the form of demonstrations on performance and robustness. These findings reveal that GPT models are still not fully proficient in handling common NLP tasks, and highlight the difficulty in addressing robustness challenges through enhancing model performance or altering prompt content. By comparing the performance and robustness of the updated version of gpt-3.5-turbo, gpt-4, LLaMA2-7B and LLaMA2-13B, we further validate the experimental findings. Future studies on large language models should strive to enhance their capacities in information extraction and semantic understanding, while simultaneously bolstering overall robustness.

, authors=Xuanting Chen1, Junjie Ye1, Can Zu1, Nuo Xu1, Tao Gui2, *, Qi Zhang1, authorsList=Xuanting Chen, Junjie Ye, Can Zu, Nuo Xu, Tao Gui, Qi Zhang, authorCompany=null, correspAuthors=Tao Gui, authorNote=

born in 1999. Master candidate. Her main research interests include natural language processing and robust models

born in 2001. PhD candidate. His main research interest includes natural language processing

born in 2000. Master candidate. Her main research interests include large language models and information extraction

born in 1998. Master candidate. Her main research interests include natural language processing and large language models

born in 1989. PhD, associate professor, master supervisor. His main research interests include pre-training models, information extraction, and robust models

born in 1981. PhD, professor, PhD supervisor. His main research interests include natural language processing and information retrieval

, correspAuthorsNote=null, copyrightStatement=Copyright © 2024 Journal of Computer Research and Development. All rights reserved., copyrightOwner=null, extLink=null, articleAbsUrl=null, sourceXml=null, magXml=null, pdfUrl=null, pdf=null, pdfFileSize=null, pdfExtLink=null, richHtmlUrl=null, mobilePdfUrl=null, reviewReport=null, pdfFirstPage=null, abstractGraph=null, abstractGraphContent=null, abstractVideo=null, citation=null, cebUrl=null, magXmlContent=null, mapNumber=null, fund=null), CN=ArticleExt(id=1329460977113870887, articleId=1329460968599433730, tenantId=1146029695717560320, journalId=1324759914146529478, language=CN, title=GPT系列大语言模型在自然语言处理任务中的鲁棒性, columnId=1329460878694527739, journalTitle=计算机研究与发展, columnName=大模型与安全专题, runingTitle=null, highlight=null, articleAbstract=

大语言模型(large language models,LLMs)所展现的处理各种自然语言处理(natural language processing,NLP)任务的能力引发了广泛关注. 然而,它们在处理现实中各种复杂场景时的鲁棒性尚未得到充分探索,这对于评估模型的稳定性和可靠性尤为重要. 因此,使用涵盖了9个常见NLP任务的15个数据集(约147000个原始测试样本)和来自TextFlint的61种鲁棒的文本变形方法分析GPT-3和GPT-3.5系列模型在原始数据集上的性能,以及其在不同任务和文本变形级别(字符、词和句子)上的鲁棒性. 研究结果表明,GPT模型虽然在情感分析、语义匹配等分类任务和阅读理解任务中表现出良好的性能,但其处理信息抽取任务的能力仍较为欠缺,比如其对关系抽取任务中各种关系类型存在严重混淆,甚至出现“幻觉”现象. 在鲁棒性评估实验中,GPT模型在任务层面和变形层面的鲁棒性都较弱,其中,在分类任务和句子级别的变形中鲁棒性缺乏更为显著. 此外,探究了模型迭代过程中性能和鲁棒性的变化,以及提示中的演示数量和演示内容对模型性能和鲁棒性的影响. 结果表明,随着模型的迭代以及上下文学习的加入,模型的性能稳步提升,但是鲁棒性依然亟待提升. 这些发现从任务类型、变形种类、提示内容等方面揭示了GPT模型还无法完全胜任常见的NLP任务,并且模型存在的鲁棒性问题难以通过提升模型性能或改变提示内容等方式解决. 通过对gpt-3.5-turbo的更新版本、gpt-4模型,以及开源模型LLaMA2-7B和LLaMA2-13B的性能和鲁棒性表现进行对比,进一步验证了实验结论. 鉴于此,未来的大模型研究应当提升模型在信息提取以及语义理解等方面的能力,并且应当在模型训练或微调阶段考虑提升其鲁棒性.

, authors=陈炫婷1, 叶俊杰1, 祖璨1, 许诺1, 桂韬2, *, 张奇1, authorsList=陈炫婷, 叶俊杰, 祖璨, 许诺, 桂韬, 张奇, authorCompany=null, correspAuthors=桂韬, authorNote=

1999年生. 硕士研究生. 主要研究方向为自然语言处理、鲁棒模型

2001年生. 博士研究生. 主要研究方向为自然语言处理

2000年生. 硕士研究生. 主要研究方向为大语言模型、信息抽取

1998年生. 硕士研究生. 主要研究方向为自然语言处理、大语言模型

1989年生. 博士,副研究员,硕士生导师. 主要研究方向为预训练模型、信息抽取、鲁棒模型

1981年生. 博士,教授,博士生导师. 主要研究方向为自然语言处理、信息检索

, correspAuthorsNote=
桂韬()
, copyrightStatement=版权所有 © 2024 计算机研究与发展编辑部, copyrightOwner=null, extLink=null, articleAbsUrl=null, sourceXml=yyGPC3dGzFDDqPbpyavWHQ==, magXml=4V2nzdQsd7e7fjfoMB6qKg==, pdfUrl=null, pdf=ZIzl9KasRmFHlfIJHkKnnA==, pdfFileSize=2030627, pdfExtLink=null, richHtmlUrl=null, mobilePdfUrl=null, reviewReport=null, pdfFirstPage=null, abstractGraph=wwirxfDbD9WI2YpJuEUFxQ==, abstractGraphContent=null, abstractVideo=null, citation=null, cebUrl=null, magXmlContent=5Azpg4C25Iet93xndQ+7zw==, mapNumber=null, fund=null)}, authors=[Author(id=1329461059645190404, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329460968599433730, orderNo=0, firstName=null, middleName=null, lastName=null, nameCn=null, orcid=null, stid=null, country=null, authorPic=null, dead=0, email=xuantingchen21@m.fudan.edu.cn, emailSecond=null, emailThird=null, correspondingAuthor=0, authorType=1, ext={EN=AuthorExt(id=1329461059708104966, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329460968599433730, authorId=1329461059645190404, language=EN, stringName=Xuanting Chen, firstName=Xuanting, middleName=null, lastName=Chen, prefix=null, suffix=null, authorComment=null, nameInitials=null, affiliation=null, department=null, xref=1, address=1School of Computer Science, Fudan University, Shanghai 200433, bio={"content":"

born in 1999. Master candidate. Her main research interests include natural language processing and robust models

"}, bioImg=null, bioContent=

born in 1999. Master candidate. Her main research interests include natural language processing and robust models

, aboutCorrespAuthor=null), CN=AuthorExt(id=1329461059758436615, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329460968599433730, authorId=1329461059645190404, language=CN, stringName=陈炫婷, firstName=null, middleName=null, lastName=null, prefix=null, suffix=null, authorComment=null, nameInitials=null, affiliation=null, department=null, xref=1, address=1复旦大学计算机科学技术学院 上海 200433, bio={"img":"gG9NFf5w/DNlLOv9X0dHOQ==","content":"

1999年生. 硕士研究生. 主要研究方向为自然语言处理、鲁棒模型

"}, bioImg=gG9NFf5w/DNlLOv9X0dHOQ==, bioContent=

1999年生. 硕士研究生. 主要研究方向为自然语言处理、鲁棒模型

, aboutCorrespAuthor=null)}, companyList=[AuthorCompany(id=1329461059498389757, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329460968599433730, xref=1, ext=[AuthorCompanyExt(id=1329461059510972670, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329460968599433730, companyId=1329461059498389757, language=EN, country=null, province=null, city=null, postcode=null, companyName=null, departmentName=null, remark=1School of Computer Science, Fudan University, Shanghai 200433), AuthorCompanyExt(id=1329461059519361279, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329460968599433730, companyId=1329461059498389757, language=CN, country=null, province=null, city=null, postcode=null, companyName=null, departmentName=null, remark=1复旦大学计算机科学技术学院 上海 200433)])]), Author(id=1329461059808768265, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329460968599433730, orderNo=1, firstName=null, middleName=null, lastName=null, nameCn=null, orcid=null, stid=null, country=null, authorPic=null, dead=0, email=null, emailSecond=null, emailThird=null, correspondingAuthor=0, authorType=1, ext={EN=AuthorExt(id=1329461059884265739, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329460968599433730, authorId=1329461059808768265, language=EN, stringName=Junjie Ye, firstName=Junjie, middleName=null, lastName=Ye, prefix=null, suffix=null, authorComment=null, nameInitials=null, affiliation=null, department=null, xref=1, address=1School of Computer Science, Fudan University, Shanghai 200433, bio={"content":"

born in 2001. PhD candidate. His main research interest includes natural language processing

"}, bioImg=null, bioContent=

born in 2001. PhD candidate. His main research interest includes natural language processing

, aboutCorrespAuthor=null), CN=AuthorExt(id=1329461059934597388, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329460968599433730, authorId=1329461059808768265, language=CN, stringName=叶俊杰, firstName=null, middleName=null, lastName=null, prefix=null, suffix=null, authorComment=null, nameInitials=null, affiliation=null, department=null, xref=1, address=1复旦大学计算机科学技术学院 上海 200433, bio={"img":"gCyKSQ+si9rUPVJIhq7jQg==","content":"

2001年生. 博士研究生. 主要研究方向为自然语言处理

"}, bioImg=gCyKSQ+si9rUPVJIhq7jQg==, bioContent=

2001年生. 博士研究生. 主要研究方向为自然语言处理

, aboutCorrespAuthor=null)}, companyList=[AuthorCompany(id=1329461059498389757, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329460968599433730, xref=1, ext=[AuthorCompanyExt(id=1329461059510972670, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329460968599433730, companyId=1329461059498389757, language=EN, country=null, province=null, city=null, postcode=null, companyName=null, departmentName=null, remark=1School of Computer Science, Fudan University, Shanghai 200433), AuthorCompanyExt(id=1329461059519361279, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329460968599433730, companyId=1329461059498389757, language=CN, country=null, province=null, city=null, postcode=null, companyName=null, departmentName=null, remark=1复旦大学计算机科学技术学院 上海 200433)])]), Author(id=1329461059989123342, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329460968599433730, orderNo=2, firstName=null, middleName=null, lastName=null, nameCn=null, orcid=null, stid=null, country=null, authorPic=null, dead=0, email=null, emailSecond=null, emailThird=null, correspondingAuthor=0, authorType=1, ext={EN=AuthorExt(id=1329461060068815120, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329460968599433730, authorId=1329461059989123342, language=EN, stringName=Can Zu, firstName=Can, middleName=null, lastName=Zu, prefix=null, suffix=null, authorComment=null, nameInitials=null, affiliation=null, department=null, xref=1, address=1School of Computer Science, Fudan University, Shanghai 200433, bio={"content":"

born in 2000. Master candidate. Her main research interests include large language models and information extraction

"}, bioImg=null, bioContent=

born in 2000. Master candidate. Her main research interests include large language models and information extraction

, aboutCorrespAuthor=null), CN=AuthorExt(id=1329461060148506897, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329460968599433730, authorId=1329461059989123342, language=CN, stringName=祖璨, firstName=null, middleName=null, lastName=null, prefix=null, suffix=null, authorComment=null, nameInitials=null, affiliation=null, department=null, xref=1, address=1复旦大学计算机科学技术学院 上海 200433, bio={"img":"qFKRjU/3w59VeYBn1lPA1Q==","content":"

2000年生. 硕士研究生. 主要研究方向为大语言模型、信息抽取

"}, bioImg=qFKRjU/3w59VeYBn1lPA1Q==, bioContent=

2000年生. 硕士研究生. 主要研究方向为大语言模型、信息抽取

, aboutCorrespAuthor=null)}, companyList=[AuthorCompany(id=1329461059498389757, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329460968599433730, xref=1, ext=[AuthorCompanyExt(id=1329461059510972670, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329460968599433730, companyId=1329461059498389757, language=EN, country=null, province=null, city=null, postcode=null, companyName=null, departmentName=null, remark=1School of Computer Science, Fudan University, Shanghai 200433), AuthorCompanyExt(id=1329461059519361279, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329460968599433730, companyId=1329461059498389757, language=CN, country=null, province=null, city=null, postcode=null, companyName=null, departmentName=null, remark=1复旦大学计算机科学技术学院 上海 200433)])]), Author(id=1329461060219810067, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329460968599433730, orderNo=3, firstName=null, middleName=null, lastName=null, nameCn=null, orcid=null, stid=null, country=null, authorPic=null, dead=0, email=null, emailSecond=null, emailThird=null, correspondingAuthor=0, authorType=1, ext={EN=AuthorExt(id=1329461060295307541, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329460968599433730, authorId=1329461060219810067, language=EN, stringName=Nuo Xu, firstName=Nuo, middleName=null, lastName=Xu, prefix=null, suffix=null, authorComment=null, nameInitials=null, affiliation=null, department=null, xref=1, address=1School of Computer Science, Fudan University, Shanghai 200433, bio={"content":"

born in 1998. Master candidate. Her main research interests include natural language processing and large language models

"}, bioImg=null, bioContent=

born in 1998. Master candidate. Her main research interests include natural language processing and large language models

, aboutCorrespAuthor=null), CN=AuthorExt(id=1329461060358222102, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329460968599433730, authorId=1329461060219810067, language=CN, stringName=许诺, firstName=null, middleName=null, lastName=null, prefix=null, suffix=null, authorComment=null, nameInitials=null, affiliation=null, department=null, xref=1, address=1复旦大学计算机科学技术学院 上海 200433, bio={"img":"lhr2AANW8jKPvfcknlUXjQ==","content":"

1998年生. 硕士研究生. 主要研究方向为自然语言处理、大语言模型

"}, bioImg=lhr2AANW8jKPvfcknlUXjQ==, bioContent=

1998年生. 硕士研究生. 主要研究方向为自然语言处理、大语言模型

, aboutCorrespAuthor=null)}, companyList=[AuthorCompany(id=1329461059498389757, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329460968599433730, xref=1, ext=[AuthorCompanyExt(id=1329461059510972670, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329460968599433730, companyId=1329461059498389757, language=EN, country=null, province=null, city=null, postcode=null, companyName=null, departmentName=null, remark=1School of Computer Science, Fudan University, Shanghai 200433), AuthorCompanyExt(id=1329461059519361279, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329460968599433730, companyId=1329461059498389757, language=CN, country=null, province=null, city=null, postcode=null, companyName=null, departmentName=null, remark=1复旦大学计算机科学技术学院 上海 200433)])]), Author(id=1329461060421136664, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329460968599433730, orderNo=4, firstName=null, middleName=null, lastName=null, nameCn=null, orcid=null, stid=null, country=null, authorPic=null, dead=0, email=tgui@fudan.edu.cn, emailSecond=null, emailThird=null, correspondingAuthor=1, authorType=1, ext={EN=AuthorExt(id=1329461060484051226, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329460968599433730, authorId=1329461060421136664, language=EN, stringName=Tao Gui, firstName=Tao, middleName=null, lastName=Gui, prefix=null, suffix=null, authorComment=null, nameInitials=null, affiliation=null, department=null, xref=2, *, address=2Institute of Modern Languages and Linguistics, Fudan University, Shanghai 200433, bio={"content":"

born in 1989. PhD, associate professor, master supervisor. His main research interests include pre-training models, information extraction, and robust models

"}, bioImg=null, bioContent=

born in 1989. PhD, associate professor, master supervisor. His main research interests include pre-training models, information extraction, and robust models

, aboutCorrespAuthor=null), CN=AuthorExt(id=1329461060567937307, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329460968599433730, authorId=1329461060421136664, language=CN, stringName=桂韬, firstName=null, middleName=null, lastName=null, prefix=null, suffix=null, authorComment=null, nameInitials=null, affiliation=null, department=null, xref=2, *, address=2复旦大学现代语言学研究院 上海 200433, bio={"img":"7ihVWeeLHwsEAeGsG3Y8PA==","content":"

1989年生. 博士,副研究员,硕士生导师. 主要研究方向为预训练模型、信息抽取、鲁棒模型

"}, bioImg=7ihVWeeLHwsEAeGsG3Y8PA==, bioContent=

1989年生. 博士,副研究员,硕士生导师. 主要研究方向为预训练模型、信息抽取、鲁棒模型

, aboutCorrespAuthor=null)}, companyList=[AuthorCompany(id=1329461059578081536, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329460968599433730, xref=2, ext=[AuthorCompanyExt(id=1329461059582275841, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329460968599433730, companyId=1329461059578081536, language=EN, country=null, province=null, city=null, postcode=null, companyName=null, departmentName=null, remark=2Institute of Modern Languages and Linguistics, Fudan University, Shanghai 200433), AuthorCompanyExt(id=1329461059590664450, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329460968599433730, companyId=1329461059578081536, language=CN, country=null, province=null, city=null, postcode=null, companyName=null, departmentName=null, remark=2复旦大学现代语言学研究院 上海 200433)])]), Author(id=1329461060630851869, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329460968599433730, orderNo=5, firstName=null, middleName=null, lastName=null, nameCn=null, orcid=null, stid=null, country=null, authorPic=null, dead=0, email=null, emailSecond=null, emailThird=null, correspondingAuthor=0, authorType=1, ext={EN=AuthorExt(id=1329461060697960735, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329460968599433730, authorId=1329461060630851869, language=EN, stringName=Qi Zhang, firstName=Qi, middleName=null, lastName=Zhang, prefix=null, suffix=null, authorComment=null, nameInitials=null, affiliation=null, department=null, xref=1, address=1School of Computer Science, Fudan University, Shanghai 200433, bio={"content":"

born in 1981. PhD, professor, PhD supervisor. His main research interests include natural language processing and information retrieval

"}, bioImg=null, bioContent=

born in 1981. PhD, professor, PhD supervisor. His main research interests include natural language processing and information retrieval

, aboutCorrespAuthor=null), CN=AuthorExt(id=1329461060769263904, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329460968599433730, authorId=1329461060630851869, language=CN, stringName=张奇, firstName=null, middleName=null, lastName=null, prefix=null, suffix=null, authorComment=null, nameInitials=null, affiliation=null, department=null, xref=1, address=1复旦大学计算机科学技术学院 上海 200433, bio={"img":"VW+ly+3/lWqhzvIypzgxxQ==","content":"

1981年生. 博士,教授,博士生导师. 主要研究方向为自然语言处理、信息检索

"}, bioImg=VW+ly+3/lWqhzvIypzgxxQ==, bioContent=

1981年生. 博士,教授,博士生导师. 主要研究方向为自然语言处理、信息检索

, aboutCorrespAuthor=null)}, companyList=[AuthorCompany(id=1329461059498389757, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329460968599433730, xref=1, ext=[AuthorCompanyExt(id=1329461059510972670, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329460968599433730, companyId=1329461059498389757, language=EN, country=null, province=null, city=null, postcode=null, companyName=null, departmentName=null, remark=1School of Computer Science, Fudan University, Shanghai 200433), AuthorCompanyExt(id=1329461059519361279, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329460968599433730, companyId=1329461059498389757, language=CN, country=null, province=null, city=null, postcode=null, companyName=null, departmentName=null, remark=1复旦大学计算机科学技术学院 上海 200433)])])], keywords=[Keyword(id=1329461060903481633, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329460968599433730, language=EN, orderNo=1, keyword=robustness), Keyword(id=1329461060958007586, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329460968599433730, language=EN, orderNo=2, keyword=GPT models), Keyword(id=1329461061020922147, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329460968599433730, language=EN, orderNo=3, keyword=large language models), Keyword(id=1329461061092225316, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329460968599433730, language=EN, orderNo=4, keyword=natural language processing), Keyword(id=1329461061150945573, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329460968599433730, language=EN, orderNo=5, keyword=reliability), Keyword(id=1329461061222248743, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329460968599433730, language=CN, orderNo=1, keyword=鲁棒性), Keyword(id=1329461061289357608, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329460968599433730, language=CN, orderNo=2, keyword=GPT模型), Keyword(id=1329461061348077865, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329460968599433730, language=CN, orderNo=3, keyword=大语言模型), Keyword(id=1329461061402603818, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329460968599433730, language=CN, orderNo=4, keyword=自然语言处理), Keyword(id=1329461061452935467, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329460968599433730, language=CN, orderNo=5, keyword=可靠性)], refs=[Reference(id=1329461064112124238, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329460968599433730, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=1, rfOrder=0, authorNames=null, journalName=null, refType=null, unstructuredReference=Wei J, Bosma M, Zhao V Y, et al. Finetuned language models are zero-shot learners[J]. arXiv preprint, arXiv: 2109.01652, 2021, articleTitle=null, refAbstract=null), Reference(id=1329461064179233103, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329460968599433730, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=2, rfOrder=1, authorNames=null, journalName=null, refType=null, unstructuredReference=Brown T B, Mann B, Ryder N, et al. Language models are few-shot learners[C/OL]//Advances in Neural Information Processing Systems. [2023-09-10].https://proceedings.neurips.cc/paper/2020/hash/1457c0d6bfcb4967418bfb8ac142f64a-Abstract.html, articleTitle=null, refAbstract=null), Reference(id=1329461064246341968, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329460968599433730, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=3, rfOrder=2, authorNames=null, journalName=null, refType=null, unstructuredReference=Touvron H, Lavril T, Izacard G, et al. LLaMA: Open and efficient foundation language models[J]. arXiv preprint, arXiv: 2302.13971, 2023, articleTitle=null, refAbstract=null), Reference(id=1329461064338616657, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329460968599433730, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=4, rfOrder=3, authorNames=null, journalName=null, refType=null, unstructuredReference=Anil R, Dai A M, Firat O, et al. PaLM 2 technical report[J]. arXiv preprint, arXiv: 2305.10403, 2023, articleTitle=null, refAbstract=null), Reference(id=1329461064401531218, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329460968599433730, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=5, rfOrder=4, authorNames=null, journalName=null, refType=null, unstructuredReference=Kaplan J, McCandlish S, Henighan T, et al. Scaling laws for neural language models[J]. arXiv preprint, arXiv: 2001.08361, 2020, articleTitle=null, refAbstract=null), Reference(id=1329461064460251475, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329460968599433730, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=6, rfOrder=5, authorNames=null, journalName=null, refType=null, unstructuredReference=Qin Chengwei, Zhang A, Zhang Zhuosheng, et al. Is ChatGPT a general-purpose natural language processing task solver?[C]//Proc of the 2023 Conf on Empirical Methods in Natural Language Processing. Stroudsburg, PA: ACL, 2023: 1339–1384, articleTitle=null, refAbstract=null), Reference(id=1329461064518971732, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329460968599433730, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=7, rfOrder=6, authorNames=null, journalName=null, refType=null, unstructuredReference=Bang Y, Cahyawijaya S, Lee N, et al. A multitask, multilingual, multimodal evaluation of chatgpt on reasoning, hallucination, and interactivity[J]. arXiv preprint, arXiv: 2302.04023, 2023, articleTitle=null, refAbstract=null), Reference(id=1329461064581886293, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329460968599433730, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=8, rfOrder=7, authorNames=null, journalName=null, refType=null, unstructuredReference=Kosinski M. Theory of mind may have spontaneously emerged in large language models[J]. arXiv preprint, arXiv: 2302.02083, 2023, articleTitle=null, refAbstract=null), Reference(id=1329461064648995158, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329460968599433730, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=9, rfOrder=8, authorNames=null, journalName=null, refType=null, unstructuredReference=Frieder S, Pinchetti L, Griffiths R R, et al. Mathematical capabilities of ChatGPT[C/OL]//Advances in Neural Information Processing Systems. [2023-09-10]. https://neurips.cc/virtual/2023/poster/73421, articleTitle=null, refAbstract=null), Reference(id=1329461064758047063, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329460968599433730, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=10, rfOrder=9, authorNames=null, journalName=null, refType=null, unstructuredReference=Wang Jindong, Hu Xixu, Hou Wenxin, et al. On the robustness of ChatGPT: An adversarial and out-of-distribution perspective[J]. arXiv preprint, arXiv: 2302.12095, 2023, articleTitle=null, refAbstract=null), Reference(id=1329461064837738840, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329460968599433730, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=11, rfOrder=10, authorNames=null, journalName=null, refType=null, unstructuredReference=Wang Boxin, Xu Chejian, Wang Shuohang, et al. Adversarial glue: A multi-task benchmark for robustness evaluation of language models[J]. arXiv preprint, arXiv: 2111.02840, 2021, articleTitle=null, refAbstract=null), Reference(id=1329461064913236313, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329460968599433730, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=12, rfOrder=11, authorNames=null, journalName=null, refType=null, unstructuredReference=Nie Y, Williams A, Dinan E, et al. Adversarial NLI: A new benchmark for natural language understanding[C]//Proc of the 58th Annual Meeting of the Association for Computational Linguistics. Stroudsburg, PA: ACL, 2020: 4885−4901, articleTitle=null, refAbstract=null), Reference(id=1329461065047454042, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329460968599433730, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=13, rfOrder=12, authorNames=null, journalName=null, refType=null, unstructuredReference=Fansi T A, Goel R, Wen Zhi, et al. DDXPlus: A new dataset for automatic medical diagnosis[C/OL]//Advances in Neural Information Processing Systems. [2023-09-10].https://proceedings.neurips.cc/paper_files/paper/2022/hash/cae73a974390c0edd95ae7aeae09139c-Abstract-Datasets_and_Benchmarks.html, articleTitle=null, refAbstract=null), Reference(id=1329461065110368603, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329460968599433730, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=14, rfOrder=13, authorNames=null, journalName=null, refType=null, unstructuredReference=Zhu Kaijie, Wang Jindong, Zhou Jiaheng, et al. PromptBench: Towards evaluating the robustness of large language models on adversarial prompts[J]. arXiv preprint, arXiv: 2306.04528, 2023, articleTitle=null, refAbstract=null), Reference(id=1329461065185866076, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329460968599433730, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=15, rfOrder=14, authorNames=null, journalName=null, refType=null, unstructuredReference=Wang Xiao, Liu Qin, Gui Tao, et al. TextFlint: Unified multilingual robustness evaluation toolkit for natural language processing[C]//Proc of the 59th Annual Meeting of the Association for Computational Linguistics and the 11th Int Joint Conf on Natural Language Processing: System Demonstrations. Stroudsburg, PA: ACL, 2021: 347−355, articleTitle=null, refAbstract=null), Reference(id=1329461065252974941, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329460968599433730, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=16, rfOrder=15, authorNames=null, journalName=null, refType=null, unstructuredReference=Dong Qingxiu, Li Lei, Dai Damai, et al. A survey for in-context learning[J]. arXiv preprint, arXiv: 2301.00234, 2022, articleTitle=null, refAbstract=null), Reference(id=1329461065349443934, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329460968599433730, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=17, rfOrder=16, authorNames=null, journalName=null, refType=null, unstructuredReference=Zhuo T Y, Huang Yujin, Chen Chunyang, et al. Exploring AI ethics of ChatGPT: A diagnostic analysis[J]. arXiv preprint, arXiv: 2301.12867, 2023, articleTitle=null, refAbstract=null), Reference(id=1329461065424941407, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329460968599433730, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=18, rfOrder=17, authorNames=null, journalName=null, refType=null, unstructuredReference=Choi J H, Hickman K E, Monahan A B, et al. ChatGPT goes to law school[J]. Journal of Legal Education, 2021, 71(3): 387, articleTitle=null, refAbstract=null), Reference(id=1329461065492050272, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329460968599433730, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=19, rfOrder=18, authorNames=null, journalName=null, refType=null, unstructuredReference=Khalil M, Er E. Will ChatGPT get you caught? Rethinking of plagiarism detection[C]//Proc of Int Conf on Human-Computer Interaction. Berlin: Springer, 2023: 475−487, articleTitle=null, refAbstract=null), Reference(id=1329461065559159137, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329460968599433730, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=20, rfOrder=19, authorNames=null, journalName=null, refType=null, unstructuredReference=Alshater M. Exploring the role of artificial intelligence in enhancing academic performance: A case study of ChatGPT[J/OL]. [2023-09-12]. http://dx. doi.org/10.2139/ssrn.4312358, articleTitle=null, refAbstract=null), Reference(id=1329461065626268002, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329460968599433730, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=21, rfOrder=20, authorNames=null, journalName=null, refType=null, unstructuredReference=Tabone W, De Winter J. Using ChatGPT for human–computer interaction research: A primer[J]. Royal Society Open Science, 2023, 10(9): 21, articleTitle=null, refAbstract=null), Reference(id=1329461065684988259, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329460968599433730, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=22, rfOrder=21, authorNames=null, journalName=null, refType=null, unstructuredReference=Jeblick K, Schachtner B, Dexl J, et al. ChatGPT makes medicine easy to swallow: An exploratory case study on simplified radiology reports[J]. European Radiology, 2023: 1−9, articleTitle=null, refAbstract=null), Reference(id=1329461065739514212, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329460968599433730, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=23, rfOrder=22, authorNames=null, journalName=null, refType=null, unstructuredReference=Biswas S. ChatGPT and the future of medical writing[J]. Radiology, 2023, 307(2): e223312, articleTitle=null, refAbstract=null), Reference(id=1329461065798234469, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329460968599433730, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=24, rfOrder=23, authorNames=null, journalName=null, refType=null, unstructuredReference=Goodfellow I J, Shlens J, Szegedy C. Explaining and harnessing adversarial examples[J]. arXiv preprint, arXiv: 1412.6572, 2014, articleTitle=null, refAbstract=null), Reference(id=1329461065865343334, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329460968599433730, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=25, rfOrder=24, authorNames=null, journalName=null, refType=null, unstructuredReference=Pontiki M, Galanis D, Pavlopoulos J, et al. SemEval-2014 Task 4: Aspect based sentiment analysis[C]//Proc of the 8th Int Workshop on Semantic Evaluation. Stroudsburg, PA: ACL, 2004: 27−35, articleTitle=null, refAbstract=null), Reference(id=1329461065936646503, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329460968599433730, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=26, rfOrder=25, authorNames=null, journalName=null, refType=null, unstructuredReference=Maas A L, Daly R E, Pham P T, et al. Learning word vectors for sentiment analysis[C]// Proc of the 49th Annual Meeting of the Association for Computational Linguistics: Human Language Technologies. Stroudsburg, PA: ACL, 2011: 142−150, articleTitle=null, refAbstract=null), Reference(id=1329461066007949672, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329460968599433730, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=27, rfOrder=26, authorNames=null, journalName=null, refType=null, unstructuredReference=Williams A, Nangia N, Bowman S R. A broad-coverage challenge corpus for sentence understanding through inference[J]. arXiv preprint, arXiv: 1704.05426, 2017, articleTitle=null, refAbstract=null), Reference(id=1329461066070864233, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329460968599433730, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=28, rfOrder=27, authorNames=null, journalName=null, refType=null, unstructuredReference=Dolan W B, Brockett C. Automatically constructing a corpus of sentential paraphrases[C]//Proc of the 3rd Int Workshop on Paraphrasing (IWP2005). Jeju Island: Asia Federation of Natural Language Processing, 2005: 9−16, articleTitle=null, refAbstract=null), Reference(id=1329461066133778794, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329460968599433730, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=29, rfOrder=28, authorNames=null, journalName=null, refType=null, unstructuredReference=Wang Zhiguo, Hamza W, Florian R. Bilateral multi-perspective matching for natural language sentences[C]//Proc of the 26th Int Joint Conf on Artificial Intelligence. Australia: IJCAI. org, 2017: 4144−4150, articleTitle=null, refAbstract=null), Reference(id=1329461066213470571, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329460968599433730, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=30, rfOrder=29, authorNames=null, journalName=null, refType=null, unstructuredReference=Levesque H, Davis E, Morgenstern L. The winograd schema challenge[C]//Proc of 13th Int Conf on the Principles of Knowledge Representation and Reasoning. Palo Alto, CA: AAAI, 2012: 552−561, articleTitle=null, refAbstract=null), Reference(id=1329461067794723180, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329460968599433730, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=31, rfOrder=30, authorNames=null, journalName=null, refType=null, unstructuredReference=Rajpurkar P, Zhang Jian, Lopyrev K, et al. SQuAD: 100, 000+ questions for machine comprehension of text[C]//Proc of the 2016 Conf on Empirical Methods in Natural Language Processing. Stroudsburg, PA: ACL, 2016: 2383−2392, articleTitle=null, refAbstract=null), Reference(id=1329461067866026349, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329460968599433730, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=32, rfOrder=31, authorNames=null, journalName=null, refType=null, unstructuredReference=Rajpurkar P, Jia R, Liang P. Know what you don’t know: Unanswerable questions for SQuAD[C]//Proc of the 56th Annual Meeting of the Association for Computational Linguistics (Volume 2: Short Papers). Stroudsburg, PA: ACL, 2018: 784−789, articleTitle=null, refAbstract=null), Reference(id=1329461067937329518, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329460968599433730, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=33, rfOrder=32, authorNames=null, journalName=null, refType=null, unstructuredReference=Marcus M, Santorini B, Marcinkiewicz M A. Building a large annotated corpus of English: The Penn Treebank[J]. Computational Linguistics, 1993, 19(2): 313−330, articleTitle=null, refAbstract=null), Reference(id=1329461068037992815, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329460968599433730, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=34, rfOrder=33, authorNames=null, journalName=null, refType=null, unstructuredReference=Sang E T K, De Meulder F. Introduction to the CoNLL-2003 shared task: Language-independent named entity recognition[C]//Proc of the 7th Conf on Natural Language Learning at HLT-NAACL 2003. Stroudsburg, PA: ACL, 2003: 142−147, articleTitle=null, refAbstract=null), Reference(id=1329461068105101680, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329460968599433730, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=35, rfOrder=34, authorNames=null, journalName=null, refType=null, unstructuredReference=Weischedel R, Palmer M, Marcus M, et al. Ontonotes release 5.0 ldc2013t19[J]. Linguistic Data Consortium, 2013, 23(1): 170, articleTitle=null, refAbstract=null), Reference(id=1329461068180599153, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329460968599433730, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=36, rfOrder=35, authorNames=null, journalName=null, refType=null, unstructuredReference=Zhang Yuhao, Zhong V, Chen Danqi, et al. Position-aware attention and supervised data improve slot filling[C]//Proc of Conf on Empirical Methods in Natural Language Processing. Stroudsburg, PA: ACL, 2017: 35−45, articleTitle=null, refAbstract=null), Reference(id=1329461068243513714, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329460968599433730, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=37, rfOrder=36, authorNames=null, journalName=null, refType=null, unstructuredReference=Ouyang L, Wu J, Jiang X, et al. Training language models to follow instructions with human feedback[C/OL]//Advances in Neural Information Processing Systems. [2023-09-10]. https://proceedings.neurips.cc/paper_files/paper/2022/hash/b1efde53be364a73914f58805a001731-Abstract-Conference.html, articleTitle=null, refAbstract=null), Reference(id=1329461068306428275, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329460968599433730, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=38, rfOrder=37, authorNames=null, journalName=null, refType=null, unstructuredReference=Christiano P F, Leike J, Brown T, et al. Deep reinforcement learning from human preferences[C/OL]// Advances in Neural Information Processing Systems. [2023-09-10]. https://papers.nips.cc/paper_files/paper/2017/hash/d5e2c0adad503c91f91df240d0cd4e49-Abstract.html, articleTitle=null, refAbstract=null), Reference(id=1329461068365148532, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329460968599433730, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=39, rfOrder=38, authorNames=null, journalName=null, refType=null, unstructuredReference=Kenton J D M W C, Toutanova L K. BERT: Pre-training of deep bidirectional Transformers for language understanding[C]//Proc of NAACL-HLT. Stroudsburg, PA: ACL, 2019: 4171−4186, articleTitle=null, refAbstract=null), Reference(id=1329461068428063093, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329460968599433730, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=40, rfOrder=39, authorNames=null, journalName=null, refType=null, unstructuredReference=Chen Lingjiao, Zaharia M, Zou J. How is ChatGPT’s behavior changing over time?[J]. arXiv preprint, arXiv: 2307.09009, 2023, articleTitle=null, refAbstract=null), Reference(id=1329461068486783350, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329460968599433730, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=41, rfOrder=40, authorNames=null, journalName=null, refType=null, unstructuredReference=Tu Shangqing, Li Chunyang, Yu Jifan, et al. ChatLog: Recording and analyzing ChatGPT across time[J]. arXiv preprint, arXiv: 2304.14106, 2023, articleTitle=null, refAbstract=null)], funds=null, companyList=[AuthorCompany(id=1329461059498389757, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329460968599433730, xref=1, ext=[AuthorCompanyExt(id=1329461059510972670, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329460968599433730, companyId=1329461059498389757, language=EN, country=null, province=null, city=null, postcode=null, companyName=null, departmentName=null, remark=1School of Computer Science, Fudan University, Shanghai 200433), AuthorCompanyExt(id=1329461059519361279, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329460968599433730, companyId=1329461059498389757, language=CN, country=null, province=null, city=null, postcode=null, companyName=null, departmentName=null, remark=1复旦大学计算机科学技术学院 上海 200433)]), AuthorCompany(id=1329461059578081536, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329460968599433730, xref=2, ext=[AuthorCompanyExt(id=1329461059582275841, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329460968599433730, companyId=1329461059578081536, language=EN, country=null, province=null, city=null, postcode=null, companyName=null, departmentName=null, remark=2Institute of Modern Languages and Linguistics, Fudan University, Shanghai 200433), AuthorCompanyExt(id=1329461059590664450, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329460968599433730, companyId=1329461059578081536, language=CN, country=null, province=null, city=null, postcode=null, companyName=null, departmentName=null, remark=2复旦大学现代语言学研究院 上海 200433)])], figs=[ArticleFig(id=1329461061633290540, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329460968599433730, language=EN, label=Figure 1, caption=The evolution of five GPT-3 and GPT-3.5 series models, figureFileSmall=bgEjcexVBAW6c78jmbUdiQ==, figureFileBig=wwirxfDbD9WI2YpJuEUFxQ==, tableContent=null), ArticleFig(id=1329461061692010797, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329460968599433730, language=CN, label=图1, caption=5个GPT-3和GPT-3.5系列模型的迭代过程, figureFileSmall=bgEjcexVBAW6c78jmbUdiQ==, figureFileBig=wwirxfDbD9WI2YpJuEUFxQ==, tableContent=null), ArticleFig(id=1329461061771702574, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329460968599433730, language=EN, label=Figure 2, caption=Overview of experimental evaluating process, figureFileSmall=uPIKPGboAoVMRyqXkL4P0Q==, figureFileBig=Maqlh10rw33TVNHZmWHmkg==, tableContent=null), ArticleFig(id=1329461061876560175, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329460968599433730, language=CN, label=图2, caption=实验评测流程图, figureFileSmall=uPIKPGboAoVMRyqXkL4P0Q==, figureFileBig=Maqlh10rw33TVNHZmWHmkg==, tableContent=null), ArticleFig(id=1329461061947863344, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329460968599433730, language=EN, label=Figure 3, caption=Performance of GPT-3.5 models and BERT, figureFileSmall=PKEqZTO3umfZUA+WDKXZ6w==, figureFileBig=RqZTFZZu7KUqlLQmvQijyQ==, tableContent=null), ArticleFig(id=1329461062019166513, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329460968599433730, language=CN, label=图3, caption=GPT-3.5模型和BERT的性能表现, figureFileSmall=PKEqZTO3umfZUA+WDKXZ6w==, figureFileBig=RqZTFZZu7KUqlLQmvQijyQ==, tableContent=null), ArticleFig(id=1329461062077886770, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329460968599433730, language=EN, label=Figure 4, caption=Distribution of prediction errors in CoNLL2003 dateset, figureFileSmall=v0IrlYqcCxvHiGkZLJxHYw==, figureFileBig=NuQ0lPxklO51hhZ2ic7RoA==, tableContent=null), ArticleFig(id=1329461062140801331, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329460968599433730, language=CN, label=图4, caption=在CoNLL2003数据集上的错误预测的分布, figureFileSmall=v0IrlYqcCxvHiGkZLJxHYw==, figureFileBig=NuQ0lPxklO51hhZ2ic7RoA==, tableContent=null), ArticleFig(id=1329461062191132980, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329460968599433730, language=EN, label=Figure 5, caption=Distribution of prediction errors in TACRED dataset, figureFileSmall=N9KPSyzVcFmOxSvcW06aPQ==, figureFileBig=V75vrhIpu/Sq5R/4Wv5B6g==, tableContent=null), ArticleFig(id=1329461062258241845, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329460968599433730, language=CN, label=图5, caption=在TACRED数据集上的错误预测的分布, figureFileSmall=N9KPSyzVcFmOxSvcW06aPQ==, figureFileBig=V75vrhIpu/Sq5R/4Wv5B6g==, tableContent=null), ArticleFig(id=1329461062308573494, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329460968599433730, language=EN, label=Figure 6, caption=Performance variations of GPT models, figureFileSmall=rzy9nMrmo87DU6ZovWweMw==, figureFileBig=xhwM7r2PUF3xA+CCioF+6Q==, tableContent=null), ArticleFig(id=1329461062363099447, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329460968599433730, language=CN, label=图6, caption=GPT模型的性能变化, figureFileSmall=rzy9nMrmo87DU6ZovWweMw==, figureFileBig=xhwM7r2PUF3xA+CCioF+6Q==, tableContent=null), ArticleFig(id=1329461062421819704, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329460968599433730, language=EN, label=Figure 7, caption=APDR variations of GPT models, figureFileSmall=0eD27LWuUioG6x/4YPYC7g==, figureFileBig=vph5UI7FeIvhZzJau6uPjg==, tableContent=null), ArticleFig(id=1329461062488928569, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329460968599433730, language=CN, label=图7, caption=GPT模型的平均性能下降率的变化, figureFileSmall=0eD27LWuUioG6x/4YPYC7g==, figureFileBig=vph5UI7FeIvhZzJau6uPjg==, tableContent=null), ArticleFig(id=1329461062543454522, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329460968599433730, language=EN, label=Figure 8, caption=Performance drop of different models on three transformation categories, figureFileSmall=wsgDyM+dDoZjrO0kUXsAWg==, figureFileBig=9Q2b02tW0AIQndIE6OY5EQ==, tableContent=null), ArticleFig(id=1329461062589591867, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329460968599433730, language=CN, label=图8, caption=不同模型在3种变形类别上的性能下降情况, figureFileSmall=wsgDyM+dDoZjrO0kUXsAWg==, figureFileBig=9Q2b02tW0AIQndIE6OY5EQ==, tableContent=null), ArticleFig(id=1329461062660895036, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329460968599433730, language=EN, label=Figure 9, caption=Original and transformed performance of GPT models on 0-shot, 1-shot, and 3-shot, figureFileSmall=0JhjS5tV2O9YmjQZsgUMUw==, figureFileBig=PZKBtyrDuD+QbfsbhHg+Hw==, tableContent=null), ArticleFig(id=1329461062728003901, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329460968599433730, language=CN, label=图9, caption=GPT模型在0-shot、1-shot、3-shot样本场景下原始性能与变形后的性能表现, figureFileSmall=0JhjS5tV2O9YmjQZsgUMUw==, figureFileBig=PZKBtyrDuD+QbfsbhHg+Hw==, tableContent=null), ArticleFig(id=1329461062799307070, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329460968599433730, language=EN, label=Figure 10, caption=APDR with original and transformed demonstrations date, figureFileSmall=xxtbHX2AxwIg+W3PytwjCA==, figureFileBig=gSkJ4/fi8wcffkLEh7Ax6Q==, tableContent=null), ArticleFig(id=1329461062862221631, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329460968599433730, language=CN, label=图10, caption=模型使用原始和变形后的演示数据的APDR, figureFileSmall=xxtbHX2AxwIg+W3PytwjCA==, figureFileBig=gSkJ4/fi8wcffkLEh7Ax6Q==, tableContent=null), ArticleFig(id=1329461062920941888, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329460968599433730, language=EN, label=Figure 11, caption=Performance of GPT and LLaMA2 models, figureFileSmall=Qws8kjb+Zhr5VLPMVgsC0w==, figureFileBig=C/i5oTIDhV0u7sInpS/puw==, tableContent=null), ArticleFig(id=1329461062983856449, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329460968599433730, language=CN, label=图11, caption=GPT和LLaMA2模型的性能表现, figureFileSmall=Qws8kjb+Zhr5VLPMVgsC0w==, figureFileBig=C/i5oTIDhV0u7sInpS/puw==, tableContent=null), ArticleFig(id=1329461063042576706, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329460968599433730, language=EN, label=Table 1, caption=

Information of 15 Datasets Used in Experiments

, figureFileSmall=null, figureFileBig=null, tableContent=
任务类型子任务类型数据集数据量评测指标
分类细粒度情感分析(ABSA)SemEval2014-Laptop[25]331准确率
SemEval2014-Restaurant[25]492准确率
情感分析(SA)IMDB[26]25000准确率
自然语言推理(NLI)MNLI-m[27]9815准确率
MNLI-mm[27]9832准确率
SNLI[27]10000准确率
语义匹配(SM)QQP[28]40430准确率
MRPC[29]1725准确率
威诺格拉德模式挑战(WSC)WSC273[30]570准确率
阅读理解机器阅读理解(MRC)SQuAD 1.1[31]9868F1
SQuAD 2.0[32]11491F1
信息抽取词性标注(POS)WSJ[33]5461准确率
命名实体识别(NER)CoNLL2003[34]3453F1
OntoNotesv5[35]4019F1
关系抽取(RE)TACRED[36]15509F1
), ArticleFig(id=1329461063126462787, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329460968599433730, language=CN, label=表1, caption=

实验使用的15个数据集的信息

, figureFileSmall=null, figureFileBig=null, tableContent=
任务类型子任务类型数据集数据量评测指标
分类细粒度情感分析(ABSA)SemEval2014-Laptop[25]331准确率
SemEval2014-Restaurant[25]492准确率
情感分析(SA)IMDB[26]25000准确率
自然语言推理(NLI)MNLI-m[27]9815准确率
MNLI-mm[27]9832准确率
SNLI[27]10000准确率
语义匹配(SM)QQP[28]40430准确率
MRPC[29]1725准确率
威诺格拉德模式挑战(WSC)WSC273[30]570准确率
阅读理解机器阅读理解(MRC)SQuAD 1.1[31]9868F1
SQuAD 2.0[32]11491F1
信息抽取词性标注(POS)WSJ[33]5461准确率
命名实体识别(NER)CoNLL2003[34]3453F1
OntoNotesv5[35]4019F1
关系抽取(RE)TACRED[36]15509F1
), ArticleFig(id=1329461063210348868, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329460968599433730, language=EN, label=Table 2, caption=

Information of 61 Task-Specific Transformations

, figureFileSmall=null, figureFileBig=null, tableContent=
子任务类型变形类型变形方式
细粒度情感分析(ABSA)句子级AddDiff, RevNon, RevTgt
情感分析(SA)词级SwapSpecialEnt-Movie,
SwapSpecialEnt-Person
句子级AddSum-Movie, AddSum-Person, DoubleDenial
自然语言推理(NLI)字符级NumWord
词级SwapAnt
句子级AddSent
语义匹配(SM)字符级NumWord
词级SwapAnt
威诺格拉德模式挑战(WSC)字符级SwapNames
词级SwapGender
句子级AddSentences, InsertRelativeClause, SwitchVoice
机器阅读理解(MRC)句子级AddSentDiverse, ModifyPos, PerturbAnswer, PerturbQuestion-BackTranslation, PertyrbQuestion-MLM
词性标注(POS)字符级SwapPrefix
词级SwapMultiPOSJJ, SwapMultiPOSNN, SwapMultiPOSRB, SwapMutliPOSVB
命名实体识别(NER)字符级EntTypos, OOV
词级CrossCategory, SwapLonger
句子级ConcatSent
关系抽取(RE)词级SwapEnt-LowFreq, SwapEnt-SamEtype
句子级InsertClause, SwapTriplePos-Age, SwapTriplePos-Birth, SwapTriplePos-Employee
), ArticleFig(id=1329461063277457733, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329460968599433730, language=CN, label=表2, caption=

61种任务特定变形的信息

, figureFileSmall=null, figureFileBig=null, tableContent=
子任务类型变形类型变形方式
细粒度情感分析(ABSA)句子级AddDiff, RevNon, RevTgt
情感分析(SA)词级SwapSpecialEnt-Movie,
SwapSpecialEnt-Person
句子级AddSum-Movie, AddSum-Person, DoubleDenial
自然语言推理(NLI)字符级NumWord
词级SwapAnt
句子级AddSent
语义匹配(SM)字符级NumWord
词级SwapAnt
威诺格拉德模式挑战(WSC)字符级SwapNames
词级SwapGender
句子级AddSentences, InsertRelativeClause, SwitchVoice
机器阅读理解(MRC)句子级AddSentDiverse, ModifyPos, PerturbAnswer, PerturbQuestion-BackTranslation, PertyrbQuestion-MLM
词性标注(POS)字符级SwapPrefix
词级SwapMultiPOSJJ, SwapMultiPOSNN, SwapMultiPOSRB, SwapMutliPOSVB
命名实体识别(NER)字符级EntTypos, OOV
词级CrossCategory, SwapLonger
句子级ConcatSent
关系抽取(RE)词级SwapEnt-LowFreq, SwapEnt-SamEtype
句子级InsertClause, SwapTriplePos-Age, SwapTriplePos-Birth, SwapTriplePos-Employee
), ArticleFig(id=1329461063449424198, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329460968599433730, language=EN, label=Table 3, caption=

Examples of Deformations in Different Categories

, figureFileSmall=null, figureFileBig=null, tableContent=
变形类型变形方式样例
字符级SwapPrefix原始:That is a prefixed string.
变形后:That is a preunfixed string.
词级DoubleDenial原始:The leading actor is good.
变形后:The leading actor is good not bad.
句子级InsertClause原始:Shanghai is in the east of China.
变形后:Shanghai which is a municipality of China
is in the east of China established in Tiananmen.
), ArticleFig(id=1329461063529115975, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329460968599433730, language=CN, label=表3, caption=

不同类型的变形样例

, figureFileSmall=null, figureFileBig=null, tableContent=
变形类型变形方式样例
字符级SwapPrefix原始:That is a prefixed string.
变形后:That is a preunfixed string.
词级DoubleDenial原始:The leading actor is good.
变形后:The leading actor is good not bad.
句子级InsertClause原始:Shanghai is in the east of China.
变形后:Shanghai which is a municipality of China
is in the east of China established in Tiananmen.
), ArticleFig(id=1329461063604613448, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329460968599433730, language=EN, label=Table 4, caption=

The Robustness Performance of Different Models %

, figureFileSmall=null, figureFileBig=null, tableContent=
数据集gpt-3.5-turbotext-davinci-003BERT
oritransAPDRoritransAPDRoritransAPDR
Restaurant91.43±1.2366.00±11.2827.80±2.7490.14±1.3352.59±11.2141.65±4.2684.38±1.2053.49±15.0736.51±18.43
Laptop86.67±2.1559.36±21.9731.25±23.3183.30±0.7154.71±17.7534.42±19.2990.48±0.0649.06±9.0345.78±9.97
IMDB91.60±0.2090.86±0.500.80±0.4791.74±0.6891.40±0.580.37±0.3195.24±0.1294.61±0.800.66±0.94
MNLI-m73.03±7.4441.75±17.0542.27±21.8767.49±2.8054.88±20.9319.52±24.6086.31±4.5052.49±2.9739.10±4.13
MNLI-mm72.21±7.6940.94±19.1142.71±24.3166.61±1.5750.57±20.5824.46±27.7184.17±1.0952.33±5.4437.87±5.73
SNLI73.30±12.5047.80±8.8032.99±13.6670.81±9.2456.44±22.6818.99±26.1690.75±1.5277.61±18.3414.44±20.25
QQP79.32±5.9764.96±20.5217.17±1.1870.14±12.0369.27±13.67−1.08±9.2391.75±2.6052.77±5.9342.56±4.83
MRPC80.69±10.2884.99±10.69−8.12±22.9974.87±5.3874.33±23.12−0.17±26.5186.87±6.050.00±0.00100.00±0.00
WSC27366.05±1.9564.12±5.822.93±5.5762.05±0.4861.42±2.411.01±3.1256.00±0.0053.61±5.314.26±9.49
SQuAD 1.155.33±8.2244.55±9.7319.45±12.3967.18±8.2361.07±9.049.11±7.1387.22±0.2670.78±21.8418.88±24.95
SQuAD 2.055.03±7.3944.21±9.3119.62±12.7065.91±7.8159.70±8.939.45±7.5878.81±2.6560.17±16.9923.48±21.81
WSJ−−−75.53±2.2874.63±2.581.21±0.9097.72±0.0996.23±1.691.53±1.79
CoNLL200344.61±3.4837.30±9.2916.31±20.0551.54±2.8842.64±9.2417.13±17.7690.57±0.3872.24±16.7520.26±18.42
OntoNotesv517.74±8.5118.68±7.00−12.73±40.0911.94±9.9812.30±7.69−17.51±51.7379.99±6.5461.98±20.3023.47±20.45
TACRED31.44±31.2432.64±33.270.58±7.8835.67±30.8938.67±31.59−25.69±55.1477.99±13.4765.53±15.4616.54±7.83
), ArticleFig(id=1329461063671722313, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329460968599433730, language=CN, label=表4, caption=

不同模型的鲁棒性表现

, figureFileSmall=null, figureFileBig=null, tableContent=
数据集gpt-3.5-turbotext-davinci-003BERT
oritransAPDRoritransAPDRoritransAPDR
Restaurant91.43±1.2366.00±11.2827.80±2.7490.14±1.3352.59±11.2141.65±4.2684.38±1.2053.49±15.0736.51±18.43
Laptop86.67±2.1559.36±21.9731.25±23.3183.30±0.7154.71±17.7534.42±19.2990.48±0.0649.06±9.0345.78±9.97
IMDB91.60±0.2090.86±0.500.80±0.4791.74±0.6891.40±0.580.37±0.3195.24±0.1294.61±0.800.66±0.94
MNLI-m73.03±7.4441.75±17.0542.27±21.8767.49±2.8054.88±20.9319.52±24.6086.31±4.5052.49±2.9739.10±4.13
MNLI-mm72.21±7.6940.94±19.1142.71±24.3166.61±1.5750.57±20.5824.46±27.7184.17±1.0952.33±5.4437.87±5.73
SNLI73.30±12.5047.80±8.8032.99±13.6670.81±9.2456.44±22.6818.99±26.1690.75±1.5277.61±18.3414.44±20.25
QQP79.32±5.9764.96±20.5217.17±1.1870.14±12.0369.27±13.67−1.08±9.2391.75±2.6052.77±5.9342.56±4.83
MRPC80.69±10.2884.99±10.69−8.12±22.9974.87±5.3874.33±23.12−0.17±26.5186.87±6.050.00±0.00100.00±0.00
WSC27366.05±1.9564.12±5.822.93±5.5762.05±0.4861.42±2.411.01±3.1256.00±0.0053.61±5.314.26±9.49
SQuAD 1.155.33±8.2244.55±9.7319.45±12.3967.18±8.2361.07±9.049.11±7.1387.22±0.2670.78±21.8418.88±24.95
SQuAD 2.055.03±7.3944.21±9.3119.62±12.7065.91±7.8159.70±8.939.45±7.5878.81±2.6560.17±16.9923.48±21.81
WSJ−−−75.53±2.2874.63±2.581.21±0.9097.72±0.0996.23±1.691.53±1.79
CoNLL200344.61±3.4837.30±9.2916.31±20.0551.54±2.8842.64±9.2417.13±17.7690.57±0.3872.24±16.7520.26±18.42
OntoNotesv517.74±8.5118.68±7.00−12.73±40.0911.94±9.9812.30±7.69−17.51±51.7379.99±6.5461.98±20.3023.47±20.45
TACRED31.44±31.2432.64±33.270.58±7.8835.67±30.8938.67±31.59−25.69±55.1477.99±13.4765.53±15.4616.54±7.83
), ArticleFig(id=1329461063751414090, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329460968599433730, language=EN, label=Table 5, caption=

The Robustness Performance of Three GPT Models %

, figureFileSmall=null, figureFileBig=null, tableContent=
数据集gpt-3.5-turbo-0301gpt-3.5-turbo-0613gpt-4
oritransAPDRoritransAPDRoritransAPDR
Restaurant91.43±1.2366.00±11.2827.80±2.7497.05±0.8659.98±16.3738.28±16.5695.81±2.2771.07±9.1525.80±9.69
Laptop86.67±2.1559.36±21.9731.25±23.3193.91±1.4563.82±19.1032.16±19.8398.74±1.8874.42±16.0124.75±15.42
IMDB91.60±0.2090.86±0.500.80±0.4796.58±1.0595.99±1.630.62±0.9093.81±3.6991.91±5.312.05±3.83
MNLI-m73.03±7.4441.75±17.0542.27±21.8771.88±7.9935.30±16.0051.85±20.0384.24±7.0053.46±10.5036.81±9.04
MNLI-mm72.21±7.6940.94±19.1142.71±24.3171.78±7.6835.59±15.4550.28±22.5080.23±8.1453.88±14.1933.28±14.43
SNLI73.30±12.5047.80±8.8032.99±13.6675.67±15.7038.58±11.1147.61±16.4089.10±5.6470.65±21.6021.25±21.31
QQP79.32±5.9764.96±20.5217.17±1.1881.42±8.4949.71±16.1638.22±22.6653.14±19.4884.91±15.74−105.86±159.05
MRPC80.69±10.2884.99±10.69−8.12±22.9985.70±11.1670.65±16.7414.29±30.4960.38±7.0694.65±4.68−58.46±18.46
WSC27366.05±1.9564.12±5.822.93±5.5753.98±0.7551.92±3.133.80±6.1077.88±6.1264.42±23.5716.91±30.39
SQuAD1.155.33±8.2244.55±9.7319.45±12.3990.11±1.0980.84±8.6510.27±9.7095.14±1.7484.96±13.7510.69±14.41
SQuAD2.055.03±7.3944.21±9.3119.62±12.7073.68±4.6164.25±10.7612.85±13.1681.94±3.1774.15±7.179.50±8.02
WSJ−−−50.35±5.2249.31±5.612.07±4.5268.66±3.0367.88±5.581.10±7.39
CoNLL200344.61±3.4837.30±9.2916.31±20.0566.78±2.9849.76±11.6925.38±17.6983.23±1.8665.53±13.8621.25±16.66
OntoNotesv517.74±8.5118.68±7.00−12.73±40.099.85±6.5313.50±4.13−66.86±72.427.58±15.726.70±10.7010.87±15.47
TACRED31.44±31.2432.64±33.270.58±7.8837.00±35.2940.23±34.38−20.07±36.3314.32±7.5713.31±9.17−0.02±74.59
), ArticleFig(id=1329461063826911563, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329460968599433730, language=CN, label=表5, caption=

3个GPT模型的鲁棒性表现

, figureFileSmall=null, figureFileBig=null, tableContent=
数据集gpt-3.5-turbo-0301gpt-3.5-turbo-0613gpt-4
oritransAPDRoritransAPDRoritransAPDR
Restaurant91.43±1.2366.00±11.2827.80±2.7497.05±0.8659.98±16.3738.28±16.5695.81±2.2771.07±9.1525.80±9.69
Laptop86.67±2.1559.36±21.9731.25±23.3193.91±1.4563.82±19.1032.16±19.8398.74±1.8874.42±16.0124.75±15.42
IMDB91.60±0.2090.86±0.500.80±0.4796.58±1.0595.99±1.630.62±0.9093.81±3.6991.91±5.312.05±3.83
MNLI-m73.03±7.4441.75±17.0542.27±21.8771.88±7.9935.30±16.0051.85±20.0384.24±7.0053.46±10.5036.81±9.04
MNLI-mm72.21±7.6940.94±19.1142.71±24.3171.78±7.6835.59±15.4550.28±22.5080.23±8.1453.88±14.1933.28±14.43
SNLI73.30±12.5047.80±8.8032.99±13.6675.67±15.7038.58±11.1147.61±16.4089.10±5.6470.65±21.6021.25±21.31
QQP79.32±5.9764.96±20.5217.17±1.1881.42±8.4949.71±16.1638.22±22.6653.14±19.4884.91±15.74−105.86±159.05
MRPC80.69±10.2884.99±10.69−8.12±22.9985.70±11.1670.65±16.7414.29±30.4960.38±7.0694.65±4.68−58.46±18.46
WSC27366.05±1.9564.12±5.822.93±5.5753.98±0.7551.92±3.133.80±6.1077.88±6.1264.42±23.5716.91±30.39
SQuAD1.155.33±8.2244.55±9.7319.45±12.3990.11±1.0980.84±8.6510.27±9.7095.14±1.7484.96±13.7510.69±14.41
SQuAD2.055.03±7.3944.21±9.3119.62±12.7073.68±4.6164.25±10.7612.85±13.1681.94±3.1774.15±7.179.50±8.02
WSJ−−−50.35±5.2249.31±5.612.07±4.5268.66±3.0367.88±5.581.10±7.39
CoNLL200344.61±3.4837.30±9.2916.31±20.0566.78±2.9849.76±11.6925.38±17.6983.23±1.8665.53±13.8621.25±16.66
OntoNotesv517.74±8.5118.68±7.00−12.73±40.099.85±6.5313.50±4.13−66.86±72.427.58±15.726.70±10.7010.87±15.47
TACRED31.44±31.2432.64±33.270.58±7.8837.00±35.2940.23±34.38−20.07±36.3314.32±7.5713.31±9.17−0.02±74.59
), ArticleFig(id=1329461063902409036, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329460968599433730, language=EN, label=Table 6, caption=

The Robustness Performance of LLaMA2 Model %

, figureFileSmall=null, figureFileBig=null, tableContent=
数据集LLaMA2-7BLLaMA2-13B
oritransAPDRoritransAPDR
Restaurant87.85±1.6852.38±7.0140.34±8.2287.10±3.1735.16±9.0759.84±9.45
Laptop79.40±2.9356.23±12.6828.96±16.8681.15±2.8247.21±18.5841.87±22.81
IMDB92.04±1.6891.06±2.681.08±1.4388.17±2.3087.40±2.890.88±1.21
MNLI-m46.76±16.0327.64±13.3934.77±34.6554.47±15.1544.70±18.9512.52±43.92
MNLI-mm50.16±17.2327.92±13.9939.21±32.2957.04±15.1145.47±19.3015.94±42.02
SNLI47.77±19.7330.73±17.4427.79±41.4354.79±15.2043.75±24.2212.83±53.93
QQP59.93±16.7733.18±11.0240.58±24.6154.49±12.9140.17±14.4521.36±32.47
MRPC70.66±14.7666.49±16.681.92±33.6269.59±17.7433.75±32.7043.09±63.48
WSC27352.40±3.6053.10±1.68−1.65±7.4852.57±0.7356.43±2.77−7.33±4.58
SQuAD1.179.64±0.6967.85±9.9814.80±12.5171.27±1.1663.67±5.1410.65±7.12
SQuAD2.078.25±0.9566.30±9.6615.26±12.3669.40±1.2761.77±5.0510.99±7.20
WSJ−−−−−−
CoNLL200320.05±8.924.44±5.3674.37±36.9345.66±10.2220.26±10.2753.47±26.94
OntoNotesv54.97±2.574.94±2.03−19.85±76.915.87±5.215.36±3.34−8.23±51.59
TACRED−−−4.26±2.605.95±5.45−16.67±104.08
), ArticleFig(id=1329461063969517901, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329460968599433730, language=CN, label=表6, caption=

LLaMA2模型的鲁棒性表现

, figureFileSmall=null, figureFileBig=null, tableContent=
数据集LLaMA2-7BLLaMA2-13B
oritransAPDRoritransAPDR
Restaurant87.85±1.6852.38±7.0140.34±8.2287.10±3.1735.16±9.0759.84±9.45
Laptop79.40±2.9356.23±12.6828.96±16.8681.15±2.8247.21±18.5841.87±22.81
IMDB92.04±1.6891.06±2.681.08±1.4388.17±2.3087.40±2.890.88±1.21
MNLI-m46.76±16.0327.64±13.3934.77±34.6554.47±15.1544.70±18.9512.52±43.92
MNLI-mm50.16±17.2327.92±13.9939.21±32.2957.04±15.1145.47±19.3015.94±42.02
SNLI47.77±19.7330.73±17.4427.79±41.4354.79±15.2043.75±24.2212.83±53.93
QQP59.93±16.7733.18±11.0240.58±24.6154.49±12.9140.17±14.4521.36±32.47
MRPC70.66±14.7666.49±16.681.92±33.6269.59±17.7433.75±32.7043.09±63.48
WSC27352.40±3.6053.10±1.68−1.65±7.4852.57±0.7356.43±2.77−7.33±4.58
SQuAD1.179.64±0.6967.85±9.9814.80±12.5171.27±1.1663.67±5.1410.65±7.12
SQuAD2.078.25±0.9566.30±9.6615.26±12.3669.40±1.2761.77±5.0510.99±7.20
WSJ−−−−−−
CoNLL200320.05±8.924.44±5.3674.37±36.9345.66±10.2220.26±10.2753.47±26.94
OntoNotesv54.97±2.574.94±2.03−19.85±76.915.87±5.215.36±3.34−8.23±51.59
TACRED−−−4.26±2.605.95±5.45−16.67±104.08
)], attaches=null, journal=Journal(id=1324759606947315908, delFlag=0, nameCn=计算机研究与发展, nameEn=Journal of Computer Research and Development, nameHistory1=null, nameHistory2=null, issn=1000-1239, eissn=null, cn=11-1777/TP, coden=null, periodic=0, language=CN, oaType=null, ccby=null, superviseOffice=null, ownerOffice=null, pubOffice=null, editorOffice=null, officeType=null, aims=null, clcCode=null, officeProv=null, officeCity=null, officeAddr=null, officeZip=null, officeEmail=null, officePhone=null, editDirector=null, officeDirector=null, officeDirectorPhone=null, officeStaffNum=null, officeEmpNum=null, coverPicUrl=k2l8zgqwstuxxevFX2qxVA==, journalPrice=null, startedYear=null, abbrevIsoEn=Journal of Computer Research and Development, journalRemark=null, publicationField=null, createdTime=1793776903141, updatedTime=1793850698612, createdBy=18614031015, updatedBy=13701087609, firstLetterCn=J, firstLetterEn=J, subjectCode=Electronic Information, subjectName=null, subjectCodeEn=Electronic Information, subjectNameEn=null, picCn=k2l8zgqwstuxxevFX2qxVA==, picEn=r6k0Ry+RD+c/VXj5QarmDw==, jcr=null, cjcr=null, exts=[JournalExt(id=1325069127710286688, language=CN, name=计算机研究与发展, nameHistory1=null, nameHistory2=null, managedBy=, sponsoredBy=, publishedBy=, editorOffice=, officeProv=null, officeCity=null, officeAddr=, officeZip=, editDirector=, officeDirector=null, officePhone=null, coverPicUrl=null, journalRemark=, submitArticleUrl=null, websiteUrl=, createdTime=1793850698640, updatedTime=1793850698640, createdBy=13701087609, updatedBy=13701087609, submissionGuidelinesUrl=, submissionAuthorUrl=https://crad.ict.ac.cn/journalx/authorLogOn.action, submissionEditorUrl=https://crad.ict.ac.cn/journalx/editorLogOn.action, submissionReviewUrl=https://crad.ict.ac.cn/journalx/expertLogOn.action, submissionCeEditorUrl=https://crad.ict.ac.cn/journalx/editorInChiefLogOn.action, submissionAeEditorUrl=, option={"copyright":""}), JournalExt(id=1325069127777395553, language=EN, name=Journal of Computer Research and Development, nameHistory1=null, nameHistory2=null, managedBy=, sponsoredBy=, publishedBy=, editorOffice=, officeProv=null, officeCity=null, officeAddr=, officeZip=, editDirector=, officeDirector=null, officePhone=null, coverPicUrl=null, journalRemark=, submitArticleUrl=null, websiteUrl=, createdTime=1793850698656, updatedTime=1793850698656, createdBy=13701087609, updatedBy=13701087609, submissionGuidelinesUrl=, submissionAuthorUrl=https://crad.ict.ac.cn/journalx/authorLogOn.action, submissionEditorUrl=https://crad.ict.ac.cn/journalx/editorLogOn.action, submissionReviewUrl=https://crad.ict.ac.cn/journalx/expertLogOn.action, submissionCeEditorUrl=https://crad.ict.ac.cn/journalx/editorInChiefLogOn.action, submissionAeEditorUrl=, option={"copyright":""})], databaseList=null, tenantJournalId=1324759914146529478, websiteList=[Website(id=1325070300139892771, webName=null, webTitle=null, webDomain=null, webCopyrigh=null, webIpcNo=null, seoTitle=null, seoKeywords=null, seoDescription=null, tenantJournalId=null, journalId=1324759914146529478, journalNameCn=null, journalNameEn=null, grayFlag=null, tenantId=1146029695717560320, platformId=null, journalGroupId=null, journalGroupNameCn=null, journalGroupNameEn=null, type=1, domain=https://castjournals.cast.org.cn/joweb/jsjyjyfz/CN, language=CN, createTime=1793850978169, createBy=18614031015, updateTime=1793851040259, updateBy=18614031015, name=计算机研究与发展-中文, tplId=1146099689490845704, title=计算机研究与发展, delFlag=0, indexPage=/home, props=[WebsiteProps(id=1325070893751349360, tenantId=1146029695717560320, journalId=null, journalGroupId=null, siteId=1325070300139892771, code=articleTextType, value=kx, createTime=1793851119697, updateTime=1793851119697, creator=18614031015, updator=18614031015), WebsiteProps(id=1325070893730377837, tenantId=1146029695717560320, journalId=null, journalGroupId=null, siteId=1325070300139892771, code=banner, value=null, createTime=1793851119692, updateTime=1793851119692, creator=18614031015, updator=18614031015), WebsiteProps(id=1325070893772320883, tenantId=1146029695717560320, journalId=null, journalGroupId=null, siteId=1325070300139892771, code=grayFlag, value=0, createTime=1793851119702, updateTime=1793851119702, creator=18614031015, updator=18614031015), WebsiteProps(id=1325070893721989228, tenantId=1146029695717560320, journalId=null, journalGroupId=null, siteId=1325070300139892771, code=logo, value=https://castjournals.cast.org.cn/joweb/jsjyjyfz/CN/file/pic?fileId=U27tDsy717YF7XSPA+Vwdw==, createTime=1793851119690, updateTime=1793851119690, creator=18614031015, updator=18614031015), WebsiteProps(id=1325070893784903797, tenantId=1146029695717560320, journalId=null, journalGroupId=null, siteId=1325070300139892771, code=minRunFlag, value=0, createTime=1793851119705, updateTime=1793851119705, creator=18614031015, updator=18614031015), WebsiteProps(id=1325070893742960751, tenantId=1146029695717560320, journalId=null, journalGroupId=null, siteId=1325070300139892771, code=picServerUrl, value=https://castjournals.cast.org.cn/joweb/jsjyjyfz/CN/file/pic, createTime=1793851119695, updateTime=1793851119695, creator=18614031015, updator=18614031015), WebsiteProps(id=1325070893780709492, tenantId=1146029695717560320, journalId=null, journalGroupId=null, siteId=1325070300139892771, code=silenceFlag, value=0, createTime=1793851119704, updateTime=1793851119704, creator=18614031015, updator=18614031015), WebsiteProps(id=1325070893738766446, tenantId=1146029695717560320, journalId=null, journalGroupId=null, siteId=1325070300139892771, code=staticResourcePath, value=https://castjournals.cast.org.cn/joweb/cast_kjdb_cn_619/, createTime=1793851119694, updateTime=1793851119694, creator=18614031015, updator=18614031015), WebsiteProps(id=1325070893755543665, tenantId=1146029695717560320, journalId=null, journalGroupId=null, siteId=1325070300139892771, code=themeColor, value=null, createTime=1793851119698, updateTime=1793851119698, creator=18614031015, updator=18614031015), WebsiteProps(id=1325070893763932274, tenantId=1146029695717560320, journalId=null, journalGroupId=null, siteId=1325070300139892771, code=themeStyle, value=null, createTime=1793851119700, updateTime=1793851119700, creator=18614031015, updator=18614031015)]), Website(id=1325070300580294698, webName=null, webTitle=null, webDomain=null, webCopyrigh=null, webIpcNo=null, seoTitle=null, seoKeywords=null, seoDescription=null, tenantJournalId=null, journalId=1324759914146529478, journalNameCn=null, journalNameEn=null, grayFlag=null, tenantId=1146029695717560320, platformId=null, journalGroupId=null, journalGroupNameCn=null, journalGroupNameEn=null, type=1, domain=https://castjournals.cast.org.cn/joweb/jsjyjyfz/EN, language=EN, createTime=1793850978274, createBy=18614031015, updateTime=1793851036638, updateBy=18614031015, name=计算机研究与发展-英文, tplId=1146101810881728533, title=Journal of Computer Research and Development, delFlag=0, indexPage=/home, props=[WebsiteProps(id=1325070925288321148, tenantId=1146029695717560320, journalId=null, journalGroupId=null, siteId=1325070300580294698, code=articleTextType, value=kx, createTime=1793851127216, updateTime=1793851127216, creator=18614031015, updator=18614031015), WebsiteProps(id=1325070925263155321, tenantId=1146029695717560320, journalId=null, journalGroupId=null, siteId=1325070300580294698, code=banner, value=null, createTime=1793851127210, updateTime=1793851127210, creator=18614031015, updator=18614031015), WebsiteProps(id=1325070925309292671, tenantId=1146029695717560320, journalId=null, journalGroupId=null, siteId=1325070300580294698, code=grayFlag, value=0, createTime=1793851127221, updateTime=1793851127221, creator=18614031015, updator=18614031015), WebsiteProps(id=1325070925254766712, tenantId=1146029695717560320, journalId=null, journalGroupId=null, siteId=1325070300580294698, code=logo, value=https://castjournals.cast.org.cn/joweb/jsjyjyfz/EN/file/pic?fileId=U27tDsy717YF7XSPA+Vwdw==, createTime=1793851127208, updateTime=1793851127208, creator=18614031015, updator=18614031015), WebsiteProps(id=1325070925326069889, tenantId=1146029695717560320, journalId=null, journalGroupId=null, siteId=1325070300580294698, code=minRunFlag, value=0, createTime=1793851127225, updateTime=1793851127225, creator=18614031015, updator=18614031015), WebsiteProps(id=1325070925279932539, tenantId=1146029695717560320, journalId=null, journalGroupId=null, siteId=1325070300580294698, code=picServerUrl, value=https://castjournals.cast.org.cn/joweb/jsjyjyfz/EN/file/pic, createTime=1793851127214, updateTime=1793851127214, creator=18614031015, updator=18614031015), WebsiteProps(id=1325070925317681280, tenantId=1146029695717560320, journalId=null, journalGroupId=null, siteId=1325070300580294698, code=silenceFlag, value=0, createTime=1793851127223, updateTime=1793851127223, creator=18614031015, updator=18614031015), WebsiteProps(id=1325070925271543930, tenantId=1146029695717560320, journalId=null, journalGroupId=null, siteId=1325070300580294698, code=staticResourcePath, value=https://castjournals.cast.org.cn/joweb/cast_kjdb_en_623/, createTime=1793851127212, updateTime=1793851127212, creator=18614031015, updator=18614031015), WebsiteProps(id=1325070925292515453, tenantId=1146029695717560320, journalId=null, journalGroupId=null, siteId=1325070300580294698, code=themeColor, value=null, createTime=1793851127217, updateTime=1793851127217, creator=18614031015, updator=18614031015), WebsiteProps(id=1325070925300904062, tenantId=1146029695717560320, journalId=null, journalGroupId=null, siteId=1325070300580294698, code=themeStyle, value=null, createTime=1793851127219, updateTime=1793851127219, creator=18614031015, updator=18614031015)])], journalTitle=计算机研究与发展, weixinUrl=null, journalUrl=https://crad.ict.ac.cn/, iacademicId=null, status=1, seqNo=null, journalTitleEn=Journal of Computer Research and Development, journalPhotoCn=k2l8zgqwstuxxevFX2qxVA==, journalPhotoEn=r6k0Ry+RD+c/VXj5QarmDw==, journalFirstLetter=J, journalRecommend=null, journalNew=null, journalCollection=null, jcrJf=null, cjcrJf=null, jcrJfStr=null, cjcrJfStr=null, submissionFirstDecision=null, sciSubjectClassification=null, casSubjectClassification=null, citeScore=null, totalCitationFrequency=null, icpCode=null, psCode=null, advertisingLicenseCode=null, copyrightInformation=null, country=null, option=, provinceCode=null, provinceName=null, collectFlag=false, interPubPlatform=, interPubPlatformUrl=null), detailUrlCn=https://castjournals.cast.org.cn/joweb/jsjyjyfz/CN/10.7544/issn1000-1239.202330801, detailUrlEn=https://castjournals.cast.org.cn/joweb/jsjyjyfz/EN/10.7544/issn1000-1239.202330801, pdfUrlCn=https://castjournals.cast.org.cn/joweb/jsjyjyfz/CN/PDF/10.7544/issn1000-1239.202330801, pdfUrlEn=https://castjournals.cast.org.cn/joweb/jsjyjyfz/EN/PDF/10.7544/issn1000-1239.202330801, aliStartDate=0, aliEndDate=0, collectionFlag=false, citedCount=null, citedUrl=null, previewStatus=0, delFlag=0, hasFullText=1, orderTime=1714492800000, fullTextJson=null, articleText=null, reference=null)
收藏切换
GPT系列大语言模型在自然语言处理任务中的鲁棒性
陈炫婷 1 , 叶俊杰 1 , 祖璨 1 , 许诺 1 , 桂韬 2, * , 张奇 1
计算机研究与发展 | 大模型与安全专题 2024,61(5): 1128-1142
收起
收藏切换
计算机研究与发展 |大模型与安全专题 2024 , 61 (5) : 1128 -1142
GPT系列大语言模型在自然语言处理任务中的鲁棒性
全屏
[Author(id=1329461059645190404, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329460968599433730, orderNo=0, firstName=null, middleName=null, lastName=null, nameCn=null, orcid=null, stid=null, country=null, authorPic=null, dead=0, email=xuantingchen21@m.fudan.edu.cn, emailSecond=null, emailThird=null, correspondingAuthor=0, authorType=1, ext={EN=AuthorExt(id=1329461059708104966, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329460968599433730, authorId=1329461059645190404, language=EN, stringName=Xuanting Chen, firstName=Xuanting, middleName=null, lastName=Chen, prefix=null, suffix=null, authorComment=null, nameInitials=null, affiliation=null, department=null, xref=1, address=1School of Computer Science, Fudan University, Shanghai 200433, bio={"content":"

born in 1999. Master candidate. Her main research interests include natural language processing and robust models

"}, bioImg=null, bioContent=

born in 1999. Master candidate. Her main research interests include natural language processing and robust models

, aboutCorrespAuthor=null), CN=AuthorExt(id=1329461059758436615, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329460968599433730, authorId=1329461059645190404, language=CN, stringName=陈炫婷, firstName=null, middleName=null, lastName=null, prefix=null, suffix=null, authorComment=null, nameInitials=null, affiliation=null, department=null, xref=1, address=1复旦大学计算机科学技术学院 上海 200433, bio={"img":"gG9NFf5w/DNlLOv9X0dHOQ==","content":"

1999年生. 硕士研究生. 主要研究方向为自然语言处理、鲁棒模型

"}, bioImg=gG9NFf5w/DNlLOv9X0dHOQ==, bioContent=

1999年生. 硕士研究生. 主要研究方向为自然语言处理、鲁棒模型

, aboutCorrespAuthor=null)}, companyList=[AuthorCompany(id=1329461059498389757, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329460968599433730, xref=1, ext=[AuthorCompanyExt(id=1329461059510972670, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329460968599433730, companyId=1329461059498389757, language=EN, country=null, province=null, city=null, postcode=null, companyName=null, departmentName=null, remark=1School of Computer Science, Fudan University, Shanghai 200433), AuthorCompanyExt(id=1329461059519361279, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329460968599433730, companyId=1329461059498389757, language=CN, country=null, province=null, city=null, postcode=null, companyName=null, departmentName=null, remark=1复旦大学计算机科学技术学院 上海 200433)])]), Author(id=1329461059808768265, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329460968599433730, orderNo=1, firstName=null, middleName=null, lastName=null, nameCn=null, orcid=null, stid=null, country=null, authorPic=null, dead=0, email=null, emailSecond=null, emailThird=null, correspondingAuthor=0, authorType=1, ext={EN=AuthorExt(id=1329461059884265739, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329460968599433730, authorId=1329461059808768265, language=EN, stringName=Junjie Ye, firstName=Junjie, middleName=null, lastName=Ye, prefix=null, suffix=null, authorComment=null, nameInitials=null, affiliation=null, department=null, xref=1, address=1School of Computer Science, Fudan University, Shanghai 200433, bio={"content":"

born in 2001. PhD candidate. His main research interest includes natural language processing

"}, bioImg=null, bioContent=

born in 2001. PhD candidate. His main research interest includes natural language processing

, aboutCorrespAuthor=null), CN=AuthorExt(id=1329461059934597388, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329460968599433730, authorId=1329461059808768265, language=CN, stringName=叶俊杰, firstName=null, middleName=null, lastName=null, prefix=null, suffix=null, authorComment=null, nameInitials=null, affiliation=null, department=null, xref=1, address=1复旦大学计算机科学技术学院 上海 200433, bio={"img":"gCyKSQ+si9rUPVJIhq7jQg==","content":"

2001年生. 博士研究生. 主要研究方向为自然语言处理

"}, bioImg=gCyKSQ+si9rUPVJIhq7jQg==, bioContent=

2001年生. 博士研究生. 主要研究方向为自然语言处理

, aboutCorrespAuthor=null)}, companyList=[AuthorCompany(id=1329461059498389757, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329460968599433730, xref=1, ext=[AuthorCompanyExt(id=1329461059510972670, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329460968599433730, companyId=1329461059498389757, language=EN, country=null, province=null, city=null, postcode=null, companyName=null, departmentName=null, remark=1School of Computer Science, Fudan University, Shanghai 200433), AuthorCompanyExt(id=1329461059519361279, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329460968599433730, companyId=1329461059498389757, language=CN, country=null, province=null, city=null, postcode=null, companyName=null, departmentName=null, remark=1复旦大学计算机科学技术学院 上海 200433)])]), Author(id=1329461059989123342, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329460968599433730, orderNo=2, firstName=null, middleName=null, lastName=null, nameCn=null, orcid=null, stid=null, country=null, authorPic=null, dead=0, email=null, emailSecond=null, emailThird=null, correspondingAuthor=0, authorType=1, ext={EN=AuthorExt(id=1329461060068815120, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329460968599433730, authorId=1329461059989123342, language=EN, stringName=Can Zu, firstName=Can, middleName=null, lastName=Zu, prefix=null, suffix=null, authorComment=null, nameInitials=null, affiliation=null, department=null, xref=1, address=1School of Computer Science, Fudan University, Shanghai 200433, bio={"content":"

born in 2000. Master candidate. Her main research interests include large language models and information extraction

"}, bioImg=null, bioContent=

born in 2000. Master candidate. Her main research interests include large language models and information extraction

, aboutCorrespAuthor=null), CN=AuthorExt(id=1329461060148506897, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329460968599433730, authorId=1329461059989123342, language=CN, stringName=祖璨, firstName=null, middleName=null, lastName=null, prefix=null, suffix=null, authorComment=null, nameInitials=null, affiliation=null, department=null, xref=1, address=1复旦大学计算机科学技术学院 上海 200433, bio={"img":"qFKRjU/3w59VeYBn1lPA1Q==","content":"

2000年生. 硕士研究生. 主要研究方向为大语言模型、信息抽取

"}, bioImg=qFKRjU/3w59VeYBn1lPA1Q==, bioContent=

2000年生. 硕士研究生. 主要研究方向为大语言模型、信息抽取

, aboutCorrespAuthor=null)}, companyList=[AuthorCompany(id=1329461059498389757, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329460968599433730, xref=1, ext=[AuthorCompanyExt(id=1329461059510972670, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329460968599433730, companyId=1329461059498389757, language=EN, country=null, province=null, city=null, postcode=null, companyName=null, departmentName=null, remark=1School of Computer Science, Fudan University, Shanghai 200433), AuthorCompanyExt(id=1329461059519361279, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329460968599433730, companyId=1329461059498389757, language=CN, country=null, province=null, city=null, postcode=null, companyName=null, departmentName=null, remark=1复旦大学计算机科学技术学院 上海 200433)])]), Author(id=1329461060219810067, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329460968599433730, orderNo=3, firstName=null, middleName=null, lastName=null, nameCn=null, orcid=null, stid=null, country=null, authorPic=null, dead=0, email=null, emailSecond=null, emailThird=null, correspondingAuthor=0, authorType=1, ext={EN=AuthorExt(id=1329461060295307541, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329460968599433730, authorId=1329461060219810067, language=EN, stringName=Nuo Xu, firstName=Nuo, middleName=null, lastName=Xu, prefix=null, suffix=null, authorComment=null, nameInitials=null, affiliation=null, department=null, xref=1, address=1School of Computer Science, Fudan University, Shanghai 200433, bio={"content":"

born in 1998. Master candidate. Her main research interests include natural language processing and large language models

"}, bioImg=null, bioContent=

born in 1998. Master candidate. Her main research interests include natural language processing and large language models

, aboutCorrespAuthor=null), CN=AuthorExt(id=1329461060358222102, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329460968599433730, authorId=1329461060219810067, language=CN, stringName=许诺, firstName=null, middleName=null, lastName=null, prefix=null, suffix=null, authorComment=null, nameInitials=null, affiliation=null, department=null, xref=1, address=1复旦大学计算机科学技术学院 上海 200433, bio={"img":"lhr2AANW8jKPvfcknlUXjQ==","content":"

1998年生. 硕士研究生. 主要研究方向为自然语言处理、大语言模型

"}, bioImg=lhr2AANW8jKPvfcknlUXjQ==, bioContent=

1998年生. 硕士研究生. 主要研究方向为自然语言处理、大语言模型

, aboutCorrespAuthor=null)}, companyList=[AuthorCompany(id=1329461059498389757, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329460968599433730, xref=1, ext=[AuthorCompanyExt(id=1329461059510972670, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329460968599433730, companyId=1329461059498389757, language=EN, country=null, province=null, city=null, postcode=null, companyName=null, departmentName=null, remark=1School of Computer Science, Fudan University, Shanghai 200433), AuthorCompanyExt(id=1329461059519361279, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329460968599433730, companyId=1329461059498389757, language=CN, country=null, province=null, city=null, postcode=null, companyName=null, departmentName=null, remark=1复旦大学计算机科学技术学院 上海 200433)])]), Author(id=1329461060421136664, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329460968599433730, orderNo=4, firstName=null, middleName=null, lastName=null, nameCn=null, orcid=null, stid=null, country=null, authorPic=null, dead=0, email=tgui@fudan.edu.cn, emailSecond=null, emailThird=null, correspondingAuthor=1, authorType=1, ext={EN=AuthorExt(id=1329461060484051226, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329460968599433730, authorId=1329461060421136664, language=EN, stringName=Tao Gui, firstName=Tao, middleName=null, lastName=Gui, prefix=null, suffix=null, authorComment=null, nameInitials=null, affiliation=null, department=null, xref=2, *, address=2Institute of Modern Languages and Linguistics, Fudan University, Shanghai 200433, bio={"content":"

born in 1989. PhD, associate professor, master supervisor. His main research interests include pre-training models, information extraction, and robust models

"}, bioImg=null, bioContent=

born in 1989. PhD, associate professor, master supervisor. His main research interests include pre-training models, information extraction, and robust models

, aboutCorrespAuthor=null), CN=AuthorExt(id=1329461060567937307, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329460968599433730, authorId=1329461060421136664, language=CN, stringName=桂韬, firstName=null, middleName=null, lastName=null, prefix=null, suffix=null, authorComment=null, nameInitials=null, affiliation=null, department=null, xref=2, *, address=2复旦大学现代语言学研究院 上海 200433, bio={"img":"7ihVWeeLHwsEAeGsG3Y8PA==","content":"

1989年生. 博士,副研究员,硕士生导师. 主要研究方向为预训练模型、信息抽取、鲁棒模型

"}, bioImg=7ihVWeeLHwsEAeGsG3Y8PA==, bioContent=

1989年生. 博士,副研究员,硕士生导师. 主要研究方向为预训练模型、信息抽取、鲁棒模型

, aboutCorrespAuthor=null)}, companyList=[AuthorCompany(id=1329461059578081536, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329460968599433730, xref=2, ext=[AuthorCompanyExt(id=1329461059582275841, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329460968599433730, companyId=1329461059578081536, language=EN, country=null, province=null, city=null, postcode=null, companyName=null, departmentName=null, remark=2Institute of Modern Languages and Linguistics, Fudan University, Shanghai 200433), AuthorCompanyExt(id=1329461059590664450, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329460968599433730, companyId=1329461059578081536, language=CN, country=null, province=null, city=null, postcode=null, companyName=null, departmentName=null, remark=2复旦大学现代语言学研究院 上海 200433)])]), Author(id=1329461060630851869, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329460968599433730, orderNo=5, firstName=null, middleName=null, lastName=null, nameCn=null, orcid=null, stid=null, country=null, authorPic=null, dead=0, email=null, emailSecond=null, emailThird=null, correspondingAuthor=0, authorType=1, ext={EN=AuthorExt(id=1329461060697960735, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329460968599433730, authorId=1329461060630851869, language=EN, stringName=Qi Zhang, firstName=Qi, middleName=null, lastName=Zhang, prefix=null, suffix=null, authorComment=null, nameInitials=null, affiliation=null, department=null, xref=1, address=1School of Computer Science, Fudan University, Shanghai 200433, bio={"content":"

born in 1981. PhD, professor, PhD supervisor. His main research interests include natural language processing and information retrieval

"}, bioImg=null, bioContent=

born in 1981. PhD, professor, PhD supervisor. His main research interests include natural language processing and information retrieval

, aboutCorrespAuthor=null), CN=AuthorExt(id=1329461060769263904, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329460968599433730, authorId=1329461060630851869, language=CN, stringName=张奇, firstName=null, middleName=null, lastName=null, prefix=null, suffix=null, authorComment=null, nameInitials=null, affiliation=null, department=null, xref=1, address=1复旦大学计算机科学技术学院 上海 200433, bio={"img":"VW+ly+3/lWqhzvIypzgxxQ==","content":"

1981年生. 博士,教授,博士生导师. 主要研究方向为自然语言处理、信息检索

"}, bioImg=VW+ly+3/lWqhzvIypzgxxQ==, bioContent=

1981年生. 博士,教授,博士生导师. 主要研究方向为自然语言处理、信息检索

, aboutCorrespAuthor=null)}, companyList=[AuthorCompany(id=1329461059498389757, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329460968599433730, xref=1, ext=[AuthorCompanyExt(id=1329461059510972670, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329460968599433730, companyId=1329461059498389757, language=EN, country=null, province=null, city=null, postcode=null, companyName=null, departmentName=null, remark=1School of Computer Science, Fudan University, Shanghai 200433), AuthorCompanyExt(id=1329461059519361279, tenantId=1146029695717560320, journalId=1324759914146529478, articleId=1329460968599433730, companyId=1329461059498389757, language=CN, country=null, province=null, city=null, postcode=null, companyName=null, departmentName=null, remark=1复旦大学计算机科学技术学院 上海 200433)])])]
陈炫婷1 , 叶俊杰1, 祖璨1, 许诺1, 桂韬2, * , 张奇1
作者信息
  • 1复旦大学计算机科学技术学院 上海 200433
  • 2复旦大学现代语言学研究院 上海 200433
通讯作者:
桂韬()
作者简介:

1999年生. 硕士研究生. 主要研究方向为自然语言处理、鲁棒模型

2001年生. 博士研究生. 主要研究方向为自然语言处理

2000年生. 硕士研究生. 主要研究方向为大语言模型、信息抽取

1998年生. 硕士研究生. 主要研究方向为自然语言处理、大语言模型

1989年生. 博士,副研究员,硕士生导师. 主要研究方向为预训练模型、信息抽取、鲁棒模型

1981年生. 博士,教授,博士生导师. 主要研究方向为自然语言处理、信息检索

Robustness of GPT Large Language Models on Natural Language Processing Tasks
Xuanting Chen1 , Junjie Ye1, Can Zu1, Nuo Xu1, Tao Gui2, * , Qi Zhang1
Affiliations
  • 1School of Computer Science, Fudan University, Shanghai 200433
  • 2Institute of Modern Languages and Linguistics, Fudan University, Shanghai 200433
  • born in 1999. Master candidate. Her main research interests include natural language processing and robust models

    born in 2001. PhD candidate. His main research interest includes natural language processing

    born in 2000. Master candidate. Her main research interests include large language models and information extraction

    born in 1998. Master candidate. Her main research interests include natural language processing and large language models

    born in 1989. PhD, associate professor, master supervisor. His main research interests include pre-training models, information extraction, and robust models

    born in 1981. PhD, professor, PhD supervisor. His main research interests include natural language processing and information retrieval

出版时间: 2024-05-01 doi: 10.7544/issn1000-1239.202330801
文章导航
收藏切换

大语言模型(large language models,LLMs)所展现的处理各种自然语言处理(natural language processing,NLP)任务的能力引发了广泛关注. 然而,它们在处理现实中各种复杂场景时的鲁棒性尚未得到充分探索,这对于评估模型的稳定性和可靠性尤为重要. 因此,使用涵盖了9个常见NLP任务的15个数据集(约147000个原始测试样本)和来自TextFlint的61种鲁棒的文本变形方法分析GPT-3和GPT-3.5系列模型在原始数据集上的性能,以及其在不同任务和文本变形级别(字符、词和句子)上的鲁棒性. 研究结果表明,GPT模型虽然在情感分析、语义匹配等分类任务和阅读理解任务中表现出良好的性能,但其处理信息抽取任务的能力仍较为欠缺,比如其对关系抽取任务中各种关系类型存在严重混淆,甚至出现“幻觉”现象. 在鲁棒性评估实验中,GPT模型在任务层面和变形层面的鲁棒性都较弱,其中,在分类任务和句子级别的变形中鲁棒性缺乏更为显著. 此外,探究了模型迭代过程中性能和鲁棒性的变化,以及提示中的演示数量和演示内容对模型性能和鲁棒性的影响. 结果表明,随着模型的迭代以及上下文学习的加入,模型的性能稳步提升,但是鲁棒性依然亟待提升. 这些发现从任务类型、变形种类、提示内容等方面揭示了GPT模型还无法完全胜任常见的NLP任务,并且模型存在的鲁棒性问题难以通过提升模型性能或改变提示内容等方式解决. 通过对gpt-3.5-turbo的更新版本、gpt-4模型,以及开源模型LLaMA2-7B和LLaMA2-13B的性能和鲁棒性表现进行对比,进一步验证了实验结论. 鉴于此,未来的大模型研究应当提升模型在信息提取以及语义理解等方面的能力,并且应当在模型训练或微调阶段考虑提升其鲁棒性.

鲁棒性  /  GPT模型  /  大语言模型  /  自然语言处理  /  可靠性

The GPT models have demonstrated impressive performance in various natural language processing (NLP) tasks. However, their robustness and abilities to handle various complexities of the open world have not yet to be well explored, which is especially crucial in assessing the stability of models and is a key aspect of trustworthy AI. In this study, we perform a comprehensive experimental analysis of GPT-3 and GPT-3.5 series models, exploring their performance and robustness using 15 datasets (about 147000 original test samples) with 61 robust probing transformations from TextFlint covering 9 popular NLP tasks. Additionally, we analyze the model’s robustness across different transformation levels, including character, word, and sentence. Our findings reveal that while GPT models exhibit competitive performance in certain tasks like sentiment analysis, semantic matching, and reading comprehension, they exhibit severe confusion regarding information extraction tasks. For instance, GPT models exhibit severe confusion in relation extraction and even exhibit “hallucination” phenomena. Moreover, they experience significant degradation in robustness in terms of tasks and transformations, especially in classification tasks and sentence-level transformations. Furthermore, we validate the impact of the quantity and the form of demonstrations on performance and robustness. These findings reveal that GPT models are still not fully proficient in handling common NLP tasks, and highlight the difficulty in addressing robustness challenges through enhancing model performance or altering prompt content. By comparing the performance and robustness of the updated version of gpt-3.5-turbo, gpt-4, LLaMA2-7B and LLaMA2-13B, we further validate the experimental findings. Future studies on large language models should strive to enhance their capacities in information extraction and semantic understanding, while simultaneously bolstering overall robustness.

robustness  /  GPT models  /  large language models  /  natural language processing  /  reliability
陈炫婷, 叶俊杰, 祖璨, 许诺, 桂韬, 张奇. GPT系列大语言模型在自然语言处理任务中的鲁棒性. 计算机研究与发展, 2024 , 61 (5) : 1128 -1142 . DOI: 10.7544/issn1000-1239.202330801
Xuanting Chen, Junjie Ye, Can Zu, Nuo Xu, Tao Gui, Qi Zhang. Robustness of GPT Large Language Models on Natural Language Processing Tasks[J]. Journal of Computer Research and Development, 2024 , 61 (5) : 1128 -1142 . DOI: 10.7544/issn1000-1239.202330801
大语言模型,如FLAN[1], GPT-3[2], LLaMA[3]和PaLM2[4]等,在对话、理解和推理方面展示了惊人的能力[5]. 在不修改模型参数的情况下,大模型可以仅通过输入合适的提示来执行各种任务. 其中,GPT系列模型因其出色的能力备受关注.
为定量评估和探究大模型的能力,已有的工作集中于评估大模型在常识和逻辑推理[6]、多语言和多模态[7]、心智理论[8]和数学[9]等方面的能力. 尽管这些工作在基准测试集上取得了很好的效果,但大模型是否具备良好的鲁棒性仍然需要进一步研究.
鲁棒性衡量了模型在面对异常情况(如噪音、扰动或故意攻击)时的稳定性,这种能力在现实场景,尤其是在自动驾驶和医学诊断等安全场景下对于大模型至关重要. 鉴于此,现有工作对大模型的鲁棒性展开了探究:Wang等人[10]从对抗性和分布外(out of distribution,OOD)的角度出发,使用现有的AdvGLUE[11]和ANLI[12]对抗基准评估ChatGPT等大模型的对抗鲁棒性,使用DDXPlus[13]医学诊断数据集等评估分布外鲁棒性;Zhu等人[14]则从提示的角度出发,提出了基于对抗性提示的鲁棒性评测基准,并对大模型在对抗提示方面的鲁棒性进行了分析. 然而,已有的研究主要使用对抗攻击策略,这对于大规模评估来说需要消耗大量的算力和时间;并且对抗样本生成的目标是通过对特定模型或数据集的原始输入进行微小的扰动,以误导模型的分类或生成结果,但这些扰动并不总是代表真实世界中的威胁和攻击方式. 此外,现有研究大多针对ChatGPT及同时期的其他大模型,对GPT系列模型迭代过程中性能和鲁棒性的变化关注较少.
鉴于此,本文选择了图1所示的5个GPT-3和GPT-3.5系列模型作为大模型的代表,通过全面的实验分析其性能和鲁棒性,以解决3个问题.
问题1:GPT模型在自然语言处理(NLP)任务的原始数据集上有何性能缺陷?
为给后续的鲁棒性评估提供基础和参考点,本文首先评估模型在原始数据集上的性能. 本文选择15个数据集(超过147000个原始测试样本),涵盖了9个常见的NLP任务,如情感分析、阅读理解和命名实体识别等,评估了GPT模型在原始数据集上的性能以及迭代过程中的性能变化. 虽然这些任务没有直接对应具体的对话场景,但它们评估了模型的潜在能力,包括理解上下文、处理不同的语言结构和捕捉微小的信息等,这些能力对于语言理解和生成系统都非常重要.
问题2:GPT模型在NLP任务上面对输入文本扰动时的鲁棒性如何?
本文首先确定评估鲁棒性的方法. 为更加真实地模拟现实世界中可能存在的噪音、扰动和攻击,本文选择了TextFlint[15]作为对输入文本进行扰动的工具. TextFlint提供了许多针对NLP任务特定的文本变形,这些变形均基于语言学进行设计,体现了实际使用语言过程中可能发生的情况,保持了变形后文本的语言合理性,能够模拟实际应用中的挑战. 本文使用了61种文本变形方法,这些变形按照粒度可以分为句子级、词级和字符级. 本文通过实验分析了GPT模型在各种任务和各个变形级别上的鲁棒性,并探究了模型迭代过程中鲁棒性的变化.
问题3:提示对GPT模型的性能和鲁棒性有何影响?
在上述2个问题中,本文从测试文本出发,通过将不同的测试样本与任务特定的提示进行拼接,评估了模型的性能和鲁棒性. 在这个问题中,本文从提示的角度出发,研究其对性能和鲁棒性的影响. 上下文学习[16](in-context learning,ICL)已经成为NLP领域的新范式,语言模型可以仅基于少量示例执行复杂任务. 基于此,本文通过改变提示中演示(demonstration)的数量或内容,探究提示对GPT模型的性能和鲁棒性的影响.
本文的定量结果和定性分析表明:
1)GPT模型在情感分析、语义匹配等分类任务和阅读理解任务中表现出较优异的性能,但在信息抽取任务中性能较差. 例如,其严重混淆了关系抽取任务中的各种关系类型,甚至出现了“幻觉”现象.
2)在处理被扰动的输入文本时,GPT模型的鲁棒性较弱,它们在分类任务和句子级别变形中鲁棒性缺乏更为显著.
3)随着GPT系列模型的迭代,其在NLP任务上的性能稳步提升,但是鲁棒性并未增强. 除情感分析任务外,模型在其余任务上的鲁棒性均未明显提升,甚至出现显著波动.
4)随着提示中演示数量的增加,GPT模型的性能提升,但模型鲁棒性仍然亟待增强;演示内容的改变可以一定程度上增强模型的抗扰动能力,但未能从根本上解决鲁棒性问题.
同时,通过对gpt-3.5-turbo的更新版本、gpt-4、开源模型LLaMA2-7B和LLaMA2-13B的表现进行评估,本文进一步验证了上述实验结论的普适性和可持续性.
近期有大量的研究集中于评估大模型在各种任务中的性能. Qin等人[6]对ChatGPT和text-davinci-003等模型在常见NLP任务上的零样本能力进行了评测,结果表明ChatGPT擅长处理推理和对话任务,但是在序列标注任务上表现欠佳;Bang等人[7]评估了ChatGPT在多任务、多语言和多模态方面的能力,发现ChatGPT在大多数任务上优于零样本学习的大模型,甚至在某些任务上优于微调模型;Zhuo等人[17]针对大模型伦理进行了评测工作. 此外,大量工作针对大模型在不同领域的能力进行了研究和讨论,包括法律领域[18]、教育领域[19-20]、人机交互领域[21]、医学领域[22]以及写作领域[23]等. 然而,这些研究主要集中在大模型的性能上,对鲁棒性的关注有限. 模型在固定的测试数据上取得较高准确率,并不能反映出其在现实场景中面对输入的文本噪音、扰动或恶意攻击时的可靠性和稳定性,因此,鲁棒性对于评估模型处理现实世界中的复杂任务的能力至关重要.
已有的关于大模型鲁棒性的工作主要集中于2个方面:对抗鲁棒性和分布外鲁棒性. 对抗鲁棒性是指模型在对抗样本上的鲁棒性表现,对抗样本[24]的生成方式为:对原始输入施加一个阈值范围内的微小扰动,使得模型的分类或生成结果发生变化. 分布外鲁棒性关注于模型的泛化性,即使用与模型训练数据存在分布偏移的数据(包括跨域或跨时间数据)进行鲁棒性评测. Wang等人[10]使用现有的AdvGLUE[11]和ANLI[12]对抗基准评估ChatGPT等大模型的对抗性鲁棒性,使用Flipkart评论和DDXPlus[13]医学诊断数据集评估分布外鲁棒性. 结果表明,尽管ChatGPT在大多的分类任务和翻译任务上展现出更优的鲁棒性,但是大模型的对抗性和分布外鲁棒性仍然较弱. Zhu等人[14]针对提示进行对抗攻击,并使用这些对抗性提示对大模型进行鲁棒性测试,结果表明大模型容易受到对抗性提示的影响. 然而,对抗样本的数据是以欺骗模型为目的而生成的,与现实场景中产生的噪音和扰动存在明显差异,并且生成对抗样本需要消耗大量算力和时间,不适合进行大规模评测. 本文通过考虑更广泛的使用场景,从输入文本的角度出发,利用任务特定的文本变形来评估大模型在每个任务中的鲁棒性表现,从而进行更全面的分析. 此外,本文关注于GPT系列的多个模型的表现,分析了它们在迭代过程中性能和鲁棒性方面的变化.
为了全面评估GPT模型在各类NLP任务上的表现,本文选取了9个常见的NLP任务,涵盖分类、阅读理解和信息抽取3个不同类别,如表1所示. 针对每个任务,本文选取了具有代表性的公开数据集进行测试,最终共包含15个不同数据集.
根据图1所示,本文主要针对5个GPT-3和GPT-3.5系列模型进行评估和分析,并对GPT-4模型在零样本场景下进行抽样测试,所有模型都通过OpenAI官方API1进行评估. 根据OpenAI官方文档的说明,text-davinci-002是基于code-davinci-002的InstructGPT[37]模型,其使用了一种监督式微调策略的方法FeedME2进行训练;text-davinci-003是text-davinci-002的改进版本,其使用近端优化策略(proximal policy optimization,PPO)算法进行训练,该算法被用于基于人类反馈的强化学习[38](reinforcement learning from human feedback, RLHF);gpt-3.5-turbo是针对聊天场景进行优化的最强大的GPT-3.5模型(本文第3~5节所使用的版本均为gpt-3.5-turbo-0301版本).
性能评测对于评估模型的能力,以及对后续的鲁棒性评估建立基准和参考至关重要. 本节对GPT系列模型在NLP任务中原始数据集上的性能表现进行了全面的评测,旨在评估它们在不同NLP任务中的表现,并分析它们有何缺陷. 同时,本节还探究了GPT系列模型在迭代过程中的性能变化.
大模型可以通过输入适当的提示或指令来执行各种任务,而无需修改任何参数. 为评估GPT模型在NLP任务中的性能,本文针对每个具体任务设计了3种不同的提示. 如图2所示,本文将提示与测试文本拼接起来作为测试样本输入模型,并获得相应的输出,通过对输出结果的定量评估来评测模型的性能.
为定量分析模型的性能,本文使用准确率(accuracy)和F1分数(F1 score)作为评估指标. 各个数据集对应的评估指标如表1所示.
由于本文实验涉及不同模型、数据集、变形类型、提示种类等多个维度,为方便后续从不同维度对结果进行统计、计算和比较,实验选取的基准模型应当在NLP研究中具有强大的性能和广泛应用,从而能够适用于本文所有评测数据集. 因此,本文选择BERT[39]作为所有数据集的统一基准模型. 对于每个数据集,本文使用在相应数据集上经过有监督微调的BERT模型. 具体而言,对于IMDB数据集和WSJ数据集,本文使用的BERT版本分别是BERT-Large-ITPT和BERT-BiLSTM-CRF. 在其他数据集中,本文均使用BERT-base-uncased作为基准模型. 此外,本节中GPT模型的测试结果均为零样本场景下的结果.
首先分析2个最新的GPT-3.5模型(即gpt-3.5-turbo和text-davinci-003模型)的性能表现, 其和BERT在15个数据集上的性能表现如图3所示,图中的数据是每个数据集在3个提示下的性能均值. 图3所示的结果表明,GPT模型的零样本性能在情感分析、语义匹配、机器阅读理解等分类任务和阅读理解任务中可以与BERT相媲美,并且在SemEval2014-Restaurant和WSC273数据集上的表现均优于BERT.
然而,GPT模型在命名实体识别(NER)和关系抽取(RE)任务上表现不佳. 为深入了解模型错误预测背后的原因,本文选择CoNLL2003和TACRED数据集作为代表,分析了错误预测的分布情况. 图4的2个分图的第1列表示在CONLL2003数据集的预测结果中,实体类型被错误预测为“非实体”类型(即“O”)的数量. 结果表明,在NER任务中,大多数错误预测来自于“O”标签与特定实体类型的混淆,这表明大模型对实体词缺乏敏感性;在RE任务中,如图5的2个分图的第1行所示,GPT模型倾向于将“无关系”实例(即“N/A”)错误分类为特定的关系类型.
需要注意的是,我们观察到在RE任务中模型存在“幻觉”现象,即模型生成了在给定文本和预定义标签空间中不存在的虚构关系. 如图5所示,“N/A”表示“无关系”,“PER”和“ORG”分别表示属于“人物”和“组织”关系类别中的关系类型集合,而 “Other”表示不属于任何预定义标签的关系集合. 如图5的最后1列所示,GPT模型在生成结果中会虚构大量的“Other”关系,而非基于提示中给出的任务特定的关系类型和语义信息. 同时,本文在IMDB二分类数据集中也观察到类似的现象,模型将许多句子分类为“中性”标签,而该标签并不属于提示中给定的标签空间.
如图6所示,本文按照OpenAI官方发布模型的时间顺序和迭代关系(图1),评测了GPT-3和GPT-3.5系列模型在迭代过程中性能的变化. 由于测试数据较多,本文按照表1所示的子任务类型进行结果展示,每个子任务的数值为其包含数据集的结果的均值. 结果表明,随着模型发布时间的推移,GPT模型在大多数NLP任务上的性能稳步提升. 其中,GPT模型在情感分析(SA)和细粒度情感分析(ABSA)任务上保持了较高的性能,并在自然语言推理(NLI)、语义匹配(SM)和威诺格拉德模式挑战(WSC273)任务上有显著的性能提升,但在NER和RE任务上的性能一直处于较低水平.
由于text-davinci-001和gpt-3.5-turbo在WSJ数据集上未能按照提示完成任务,因此图3、图6中未展示该数据集的结果.
在NLP中,鲁棒性通常是指模型在面对噪音、扰动或有意攻击等情况时能够持续可靠地执行任务的能力. 具有较高鲁棒性的模型,在处理不应该对输出造成影响的微小变化的输入时,模型的预测结果不会发生变化. 本节对GPT模型面对输入文本扰动时的鲁棒性进行了全面评估,并分析了不同任务和不同变形级别的鲁棒性情况.
如表2所示,本节使用TextFlint提供的61种任务特定的变形来评测模型的鲁棒性. 如图2所示,每种变形均已通过TextFlint提供的变形规则作用于原始数据,从而生成变形数据. 本文通过将提示与变形数据拼接起来,作为测试文本输入模型并获得相应输出.
TextFlint提供的变形是基于语言学并针对不同的NLP任务设计的,在保持变形文本的可接受性的同时,能够更好地代表实际应用中的挑战. 本节中,根据变形的粒度,将变形分为句子级别、词级别和字符级别. 表3展示了不同类型的变形样例.
由于在不同任务和变形中使用的评估指标存在差异,本节在鲁棒性评估中引入一个新指标,即性能下降率(performance drop rate,PDR). 该指标的计算方式为:
$ PDR(T,P,{f}_{\theta },\mathcal{D})=1-\frac{\sum\limits_{(x;y)\in \mathcal{D}}\mathcal{M}\left[{f}_{\theta }([P,T(x)]),y\right]}{\sum\limits_{(x;y)\in \mathcal{D}}\mathcal{M}\left[{f}_{\theta }([P,x]),y\right]} \text{,} $
其中,$ \mathcal{M} $表示不同数据集$ \mathcal{D} $使用的评价指标. PDR提供了一种上下文归一化的度量方式,用于量化在处理经过变形T的输入$ x $(使用提示P)时,模型$ {f}_{\mathrm{\theta }} $发生的相对性能下降. 其中,负值的PDR表示在某些文本变形下会出现性能提升.
本节计算模型在不同数据集和变形中的平均原始性能(ori)、 平均变形性能(trans)和平均性能下降率(APDR). 此外,使用BERT作为基准模型,并且对于每个数据集,GPT模型和BERT都在相同的变形方法和测试数据上进行了评估.
表4列出了模型在每个数据集上的平均结果. 具体而言,本文定义$ APD{R}_{D} $为PDR(式(1))在不同数据集上的平均值:
$ {A}{P}{{D}}{R}_{{D}}(f_{\theta},{D})=\frac{1}{\left|\mathcal{T}_{{D}}\right|}\frac{1}{\left|\mathcal{P}\right|}\sum_{T\in\mathcal{T}_{{D}}}^{ }\sum_{P\in\mathcal{P}}^{ }P{D}R(T,P,f_{\theta},{D})\text{,} $
其中,$ \mathcal{T}_{{D}} $表示特定数据集D包含的任务特定变形的集合,$ \mathcal{P} $表示3个提示的集合.
与第3节类似,本节首先分析gpt-3.5-turbo和text-davinci-003的鲁棒性表现. 表4表明,GPT模型的表现与BERT类似,其在分类任务中出现了显著的性能下降. 例如,gpt-3.5-turbo在MNLI-mm数据集上的绝对性能下降了42.71个百分点,而text-davinci-003在SemEval2014-Restaurant数据集上的绝对性能下降了41.65个百分点.
此外,GPT模型在阅读理解(MRC)任务中性能较稳定,其在SQuAD 1.1和SQuAD 2.0变形前后的数据集上的性能没有出现严重的下降. 但与其他任务不同的是,在MRC任务中,text-davinci-003在性能和鲁棒性方面的表现均优于gpt-3.5-turbo. 进一步分析发现,如表4所示,gpt-3.5-turbo在该任务上具有较低的精确度(precision),通过抽样分析其生成结果,我们发现原因可能在于gpt-3.5-turbo倾向于生成更长的句子. 此外,这2个模型的输出均达到95%左右的召回率(recall),这表明GPT模型在篇章级别的理解任务上具有较强的能力.
同时,GPT模型对数字和反义词敏感度较高. 在语义匹配任务(包括QQP和MRPC数据集)中,GPT模型和BERT在变形前后的性能变化上存在显著差距. BERT在MRPC数据集上的变形后性能降至0,但GPT模型在该数据集上的变形后性能甚至有所提升. 通过分析MRPC和QQP数据集的任务特定变形,即NumWord和SwapAnt,我们发现这2种变形通过改变原始数据中的数字或对原始词语进行反义词替换,将原始句子对之间的蕴涵关系转化为矛盾关系. GPT模型在此类变形上的性能提升表明它们能够较好地捕捉到变形后的文本中数字或反义词所涉及的矛盾关系.
在NER和RE任务中,GPT模型性能的下降不明显,有时甚至有提升,尤其是在OntoNotesv5和TACRED数据集中. 但需要注意的是,模型在这些数据集上的原始性能较低. 因此,在这种情况下,讨论GPT模型在这类任务上的鲁棒性缺乏实际意义,提升模型在原始数据上的性能更为紧要.
此外,随着迭代的进行,GPT系列模型在不同任务上平均性能下降率的变化如图7所示. 由于不同模型间的结果波动较大,图7的纵坐标数值为经过对数变换之后的结果. 平均性能下降率越小,代表模型的鲁棒性越好,但图中的结果没有呈现出一致的趋势. 在ABSA和MRC任务中,模型间的鲁棒性表现较为相似;在SA任务上出现了较显著的鲁棒性提升;但是在其余任务中均呈现出显著的波动,并且没有出现鲁棒性显著提升的情况. 这可能表明GPT模型的迭代过程主要集中于改进模型在一般场景下的性能,而非解决鲁棒性问题.
图8为GPT模型在3种变形级别上的性能下降情况. 其中斜杠部分表示模型的变形后性能,无斜杠部分表示变形后性能与原始性能的差值,折线表示平均性能下降率(APDR). 通过计算每个变形级别下的PDR的均值得到$ {{A}{P}{D}{R}}_{{\mathcal{T}}_{t}} $:
$ {{APDR}}_{{\mathcal{T}}_{t}}\left({f}_{\theta },{\mathcal{T}}_{t}\right)=\frac{1}{\left|\mathcal{D}\right|}\frac{1}{\left|\mathcal{P}\right|}\sum _{D\in \mathcal{D}}\sum _{P\in \mathcal{P}}PDR\left({\mathcal{T}}_{t},P,{f}_{\theta },D\right) \text{,} $
其中,$ \mathcal{T}_{\mathcal{\mathit{t}}} $表示某个变形类别$ t $的变形集合,$ \mathcal{P} $表示提示的集合.
根据图8所示,GPT模型的APDR在句子级、词级、字符级3个变形类别上逐级递减,即处理句子级别的变形文本时,GPT模型在变形前后的性能下降更为显著. 句子级别的变形通常涉及语义的重新表述或句子整体结构的改变,这对模型稳定性有更高的要求. 此外,GPT模型在字符级和词级变形上表现出比BERT更好的鲁棒性. GPT模型的平均性能下降范围为9.61%~15.22%,而BERT在字符级和词级变形上的性能下降分别为36.74%和37.07%. 可以看出,与监督微调模型相比,GPT模型对细粒度扰动表现出更强的稳定性.
在第3节和第4节中,本文使用涵盖了各种任务和文本变形的大量测试数据,对GPT模型的性能和鲁棒性进行了评估. 除测试文本之外,提示是评测过程中模型输入数据的另一个重要部分,并且基于提示中少量示例的上下文学习已经成为NLP领域的新范式. 基于此,本节探究提示对GPT模型的性能和鲁棒性的影响,具体关注2个方面:1)提示中演示数量的影响;2)提示中演示内容的影响. 其中,演示是指提示中的示例或样本,通常用来说明我们所期望模型输出的结果.
通过改变演示数量(即图2中的“$ k $”),本文研究了在0、1和3个演示数量下模型的原始性能表现和变形前后性能的变化.
图9结果表明,增加演示数量通常会带来性能的提升. 此外,从零样本增加为少样本的情况下,模型性能提升显著,特别是对于一开始在零样本情景下表现不佳的任务,如信息抽取任务. 此外,随着演示数量的增加,不同GPT模型之间的性能差异减小.
然而,就变形前后的性能变化而言,在大多数情况下,增加演示数量没有显著缓解模型的性能下降. 只有在分类任务中,可以观察到text-davinci-001,code-davinci-002和text-davinci-002的性能下降有所缓解. 这表明增加演示数量虽然可以改善模型在原始任务上的性能,但并不能有效提高模型面对扰动时的鲁棒性.
在5.1节中的少样本情景下,原始数据和变形后数据均使用相同的、未经过变形的演示样例来研究变形后测试数据引起的性能变化. 本节研究在提示中使用变形后的演示样例对模型的鲁棒性有何影响. 本文分别从分类、信息抽取和阅读理解三大类任务中选取SemEval2014-Restaurant (Restaurant),CoNLL2003和SQuAD 1.1数据集作为代表进行实验. 对于每个数据集,演示样例使用该数据集特定的任务变形进行变换,并与变形后的测试数据拼接,用以评估模型变形后的性能. 演示样例的数量为3.
图10展示了变形前后模型的APDR. 结果表明,在演示中使用变形后的样本有助于缓解模型变形后的性能下降,说明演示中包含的扰动信息能够帮助模型更好地处理变形数据. 但是,APDR依然处于较高的数值,这表明这种性能改善是有限的,不足以从根本上解决模型的鲁棒性问题.
本文前文主要针对GPT-3和GPT-3.5系列模型的性能和鲁棒性表现进行了探究. 随着时间的推进,GPT系列模型仍然在持续迭代,并且Chen等人[40]、Tu等人[41]近期的工作表明模型的表现会随时间发生变化. 为了更好地验证本文实验结果的可持续性,本节针对GPT系列模型的更新版本“gpt-3.5-turbo-0613”(上文中的“gpt-3.5-turbo”为“gpt-3.5-turbo-0301”版本)、“gpt-4” 进行性能和鲁棒性评测.
首先是模型的性能表现. 如图11所示,根据模型更新与迭代顺序,gpt-3.5-turbo-0613和gpt-4模型在大部分数据集上的性能表现较为显著的提升. 其中,在情感分析和阅读理解的数据集中,这2个模型的提升最为显著. 第3节中的结果表明GPT模型在NER和RE任务上表现不佳,图11表明gpt-3.5-turbo-0613和gpt-4模型在NER任务的OntoNotesv5数据集及RE任务的TACRED数据集上的表现仍然处于较低水平.
其次是模型的鲁棒性表现. 表5展示了3个模型的鲁棒性表现. 如表5所示, GPT模型仍然存在4.3节中提到的鲁棒性问题,尤其在分类任务中存在显著的性能下降. 值得注意的是,在阅读理解任务中gpt-3.5-turbo-0613和gpt-4 的鲁棒性进一步提升,表现出在该任务上较高的稳定性. 同时,gpt-3.5-turbo的版本迭代未带来稳定的鲁棒性提升,而gpt-4的鲁棒性在大多任务上都优于GPT-3.5系列模型.
由于GPT系列模型出色的性能和较完善的迭代过程,对其进行的性能和鲁棒性评测有助于更全面地了解大模型的能力及其发展进程中的变化,但是由于闭源模型的限制,后续在GPT系列模型上进行优化较为困难. 为此,本节对开源模型LLaMA2-7B和LLaMA2-13B进行性能和鲁棒性评测.
如图11第1个子图所示,LLaMA2-7B和LLaMA2-13B在情感分析和阅读理解类任务上的表现与GPT-3.5系列模型相当;在第2个子图中,其在自然语言推理和语义匹配任务中却与GPT-3.5系列模型存在较大差距. 需要注意的是,LLaMA2-7B和LLaMA2-13B在WSJ和TACRED数据集中均未按照指令完成相应任务,并且在NER任务中的表现亟待提升.
如表6所示,与GPT系列模型的鲁棒性表现类似,LLaMA2-7B和LLaMA2-13B在大多分类任务上的性能下降都较为严重,但在阅读理解任务中的鲁棒性与gpt-4相当,且好于GPT-3.5系列模型. 同时,LLaMA2-13B比LLaMA2-7B具有更好的鲁棒性.
本文通过评估涵盖9个不同NLP任务的15个数据集,使用61种任务特定的变形方法,对GPT-3和GPT-3.5系列模型的性能和鲁棒性进行了全面分析. 研究结果表明,尽管GPT模型在情感分析、语义匹配等分类任务和阅读理解任务表现出色,但在面对输入文本扰动时仍然存在明显的鲁棒性问题. 其中,本文分别从任务层面和变形级别层面具体分析了GPT模型的鲁棒性表现,表明其在分类任务和句子级变形中的鲁棒性亟待提升. 同时,随着GPT系列模型的迭代,其性能在大多数任务上稳步提升,但鲁棒性依然面临很大的挑战. 此外,本文探讨了提示对GPT模型的性能和鲁棒性的影响,包括提示中演示数量和演示内容2方面. 这些发现从任务类型、变形种类、提示内容等方面揭示了 GPT模型还无法完全胜任常见的 NLP任务,并且模型存在的鲁棒性问题难以通过提升模型性能或改变提示内容等方式解决. 与此同时,本文通过评估gpt-3.5-turbo的更新版本、gpt-4模型,以及开源模型LLaMA2-7B和LLaMA2-13B的性能和鲁棒性表现,进一步验证了实验结论. 鉴于此,未来的大模型研究应当提升模型在信息提取和语义理解方面的能力,并且应当在模型训练或微调阶段考虑提升模型的鲁棒性.
作者贡献声明:陈炫婷提出研究思路和实验方案,负责部分实验和论文写作;叶俊杰负责部分实验和完善论文;祖璨负责部分实验并整理分析实验结果;许诺协助实验和完善论文;桂韬提出指导意见并修改论文;张奇提出指导意见并审阅论文.

参考文献 引证文献
排序方式:
1
Wei J, Bosma M, Zhao V Y, et al. Finetuned language models are zero-shot learners[J]. arXiv preprint, arXiv: 2109.01652, 2021
2
Brown T B, Mann B, Ryder N, et al. Language models are few-shot learners[C/OL]//Advances in Neural Information Processing Systems. [2023-09-10].https://proceedings.neurips.cc/paper/2020/hash/1457c0d6bfcb4967418bfb8ac142f64a-Abstract.html
3
Touvron H, Lavril T, Izacard G, et al. LLaMA: Open and efficient foundation language models[J]. arXiv preprint, arXiv: 2302.13971, 2023
4
Anil R, Dai A M, Firat O, et al. PaLM 2 technical report[J]. arXiv preprint, arXiv: 2305.10403, 2023
5
Kaplan J, McCandlish S, Henighan T, et al. Scaling laws for neural language models[J]. arXiv preprint, arXiv: 2001.08361, 2020
6
Qin Chengwei, Zhang A, Zhang Zhuosheng, et al. Is ChatGPT a general-purpose natural language processing task solver?[C]//Proc of the 2023 Conf on Empirical Methods in Natural Language Processing. Stroudsburg, PA: ACL, 2023: 1339–1384
7
Bang Y, Cahyawijaya S, Lee N, et al. A multitask, multilingual, multimodal evaluation of chatgpt on reasoning, hallucination, and interactivity[J]. arXiv preprint, arXiv: 2302.04023, 2023
8
Kosinski M. Theory of mind may have spontaneously emerged in large language models[J]. arXiv preprint, arXiv: 2302.02083, 2023
9
Frieder S, Pinchetti L, Griffiths R R, et al. Mathematical capabilities of ChatGPT[C/OL]//Advances in Neural Information Processing Systems. [2023-09-10]. https://neurips.cc/virtual/2023/poster/73421
10
Wang Jindong, Hu Xixu, Hou Wenxin, et al. On the robustness of ChatGPT: An adversarial and out-of-distribution perspective[J]. arXiv preprint, arXiv: 2302.12095, 2023
11
Wang Boxin, Xu Chejian, Wang Shuohang, et al. Adversarial glue: A multi-task benchmark for robustness evaluation of language models[J]. arXiv preprint, arXiv: 2111.02840, 2021
12
Nie Y, Williams A, Dinan E, et al. Adversarial NLI: A new benchmark for natural language understanding[C]//Proc of the 58th Annual Meeting of the Association for Computational Linguistics. Stroudsburg, PA: ACL, 2020: 4885−4901
13
Fansi T A, Goel R, Wen Zhi, et al. DDXPlus: A new dataset for automatic medical diagnosis[C/OL]//Advances in Neural Information Processing Systems. [2023-09-10].https://proceedings.neurips.cc/paper_files/paper/2022/hash/cae73a974390c0edd95ae7aeae09139c-Abstract-Datasets_and_Benchmarks.html
14
Zhu Kaijie, Wang Jindong, Zhou Jiaheng, et al. PromptBench: Towards evaluating the robustness of large language models on adversarial prompts[J]. arXiv preprint, arXiv: 2306.04528, 2023
15
Wang Xiao, Liu Qin, Gui Tao, et al. TextFlint: Unified multilingual robustness evaluation toolkit for natural language processing[C]//Proc of the 59th Annual Meeting of the Association for Computational Linguistics and the 11th Int Joint Conf on Natural Language Processing: System Demonstrations. Stroudsburg, PA: ACL, 2021: 347−355
16
Dong Qingxiu, Li Lei, Dai Damai, et al. A survey for in-context learning[J]. arXiv preprint, arXiv: 2301.00234, 2022
17
Zhuo T Y, Huang Yujin, Chen Chunyang, et al. Exploring AI ethics of ChatGPT: A diagnostic analysis[J]. arXiv preprint, arXiv: 2301.12867, 2023
18
Choi J H, Hickman K E, Monahan A B, et al. ChatGPT goes to law school[J]. Journal of Legal Education, 2021, 71(3): 387
19
Khalil M, Er E. Will ChatGPT get you caught? Rethinking of plagiarism detection[C]//Proc of Int Conf on Human-Computer Interaction. Berlin: Springer, 2023: 475−487
20
Alshater M. Exploring the role of artificial intelligence in enhancing academic performance: A case study of ChatGPT[J/OL]. [2023-09-12]. http://dx. doi.org/10.2139/ssrn.4312358
21
Tabone W, De Winter J. Using ChatGPT for human–computer interaction research: A primer[J]. Royal Society Open Science, 2023, 10(9): 21
22
Jeblick K, Schachtner B, Dexl J, et al. ChatGPT makes medicine easy to swallow: An exploratory case study on simplified radiology reports[J]. European Radiology, 2023: 1−9
23
Biswas S. ChatGPT and the future of medical writing[J]. Radiology, 2023, 307(2): e223312
24
Goodfellow I J, Shlens J, Szegedy C. Explaining and harnessing adversarial examples[J]. arXiv preprint, arXiv: 1412.6572, 2014
25
Pontiki M, Galanis D, Pavlopoulos J, et al. SemEval-2014 Task 4: Aspect based sentiment analysis[C]//Proc of the 8th Int Workshop on Semantic Evaluation. Stroudsburg, PA: ACL, 2004: 27−35
26
Maas A L, Daly R E, Pham P T, et al. Learning word vectors for sentiment analysis[C]// Proc of the 49th Annual Meeting of the Association for Computational Linguistics: Human Language Technologies. Stroudsburg, PA: ACL, 2011: 142−150
27
Williams A, Nangia N, Bowman S R. A broad-coverage challenge corpus for sentence understanding through inference[J]. arXiv preprint, arXiv: 1704.05426, 2017
28
Dolan W B, Brockett C. Automatically constructing a corpus of sentential paraphrases[C]//Proc of the 3rd Int Workshop on Paraphrasing (IWP2005). Jeju Island: Asia Federation of Natural Language Processing, 2005: 9−16
29
Wang Zhiguo, Hamza W, Florian R. Bilateral multi-perspective matching for natural language sentences[C]//Proc of the 26th Int Joint Conf on Artificial Intelligence. Australia: IJCAI. org, 2017: 4144−4150
30
Levesque H, Davis E, Morgenstern L. The winograd schema challenge[C]//Proc of 13th Int Conf on the Principles of Knowledge Representation and Reasoning. Palo Alto, CA: AAAI, 2012: 552−561
31
Rajpurkar P, Zhang Jian, Lopyrev K, et al. SQuAD: 100, 000+ questions for machine comprehension of text[C]//Proc of the 2016 Conf on Empirical Methods in Natural Language Processing. Stroudsburg, PA: ACL, 2016: 2383−2392
32
Rajpurkar P, Jia R, Liang P. Know what you don’t know: Unanswerable questions for SQuAD[C]//Proc of the 56th Annual Meeting of the Association for Computational Linguistics (Volume 2: Short Papers). Stroudsburg, PA: ACL, 2018: 784−789
33
Marcus M, Santorini B, Marcinkiewicz M A. Building a large annotated corpus of English: The Penn Treebank[J]. Computational Linguistics, 1993, 19(2): 313−330
34
Sang E T K, De Meulder F. Introduction to the CoNLL-2003 shared task: Language-independent named entity recognition[C]//Proc of the 7th Conf on Natural Language Learning at HLT-NAACL 2003. Stroudsburg, PA: ACL, 2003: 142−147
35
Weischedel R, Palmer M, Marcus M, et al. Ontonotes release 5.0 ldc2013t19[J]. Linguistic Data Consortium, 2013, 23(1): 170
36
Zhang Yuhao, Zhong V, Chen Danqi, et al. Position-aware attention and supervised data improve slot filling[C]//Proc of Conf on Empirical Methods in Natural Language Processing. Stroudsburg, PA: ACL, 2017: 35−45
37
Ouyang L, Wu J, Jiang X, et al. Training language models to follow instructions with human feedback[C/OL]//Advances in Neural Information Processing Systems. [2023-09-10]. https://proceedings.neurips.cc/paper_files/paper/2022/hash/b1efde53be364a73914f58805a001731-Abstract-Conference.html
38
Christiano P F, Leike J, Brown T, et al. Deep reinforcement learning from human preferences[C/OL]// Advances in Neural Information Processing Systems. [2023-09-10]. https://papers.nips.cc/paper_files/paper/2017/hash/d5e2c0adad503c91f91df240d0cd4e49-Abstract.html
39
Kenton J D M W C, Toutanova L K. BERT: Pre-training of deep bidirectional Transformers for language understanding[C]//Proc of NAACL-HLT. Stroudsburg, PA: ACL, 2019: 4171−4186
40
Chen Lingjiao, Zaharia M, Zou J. How is ChatGPT’s behavior changing over time?[J]. arXiv preprint, arXiv: 2307.09009, 2023
41
Tu Shangqing, Li Chunyang, Yu Jifan, et al. ChatLog: Recording and analyzing ChatGPT across time[J]. arXiv preprint, arXiv: 2304.14106, 2023
2024年第61卷第5期
93
45
0
引用本文
BibTeX
文章信息
doi: 10.7544/issn1000-1239.202330801
  • 接收时间:2023-10-10
  • 首发时间:2026-11-17
  • 出版时间:2024-05-01
补充材料
相关文章
文章信息
作者
出版历史
  • 收稿日期:2023-10-10
  • 修回日期:2024-03-12
基金
作者信息
    1复旦大学计算机科学技术学院 上海 200433
    2复旦大学现代语言学研究院 上海 200433

通讯作者:

桂韬()
参考文献
分享链接
https://castjournals.cast.org.cn/joweb/jsjyjyfz/CN/10.7544/issn1000-1239.202330801
分享至
全文二维码

扫描看全文

引用本文
BibTeX
本文的引用情况
2种不同金属材料的力学参数
科
Family
属数
Number of
genus
种数
Number of
species
占总种数比例
Percentage of
total species (%)
属
Genus
种数
Number of
species
占总种数比例
Percentage of total
species (%)
鹅膏菌科Amanitaceae 2 11 5.26 鹅膏菌属 Amanita 10 4.78
小菇科 Mycenaceae 2 12 5.74 丝盖伞属 Inocybe 5 2.39
多孔菌科 Polyporaceae 8 14 6.70 蜡蘑属 Laccaria 5 2.39
红菇科 Russulaceae 3 23 11.00 小皮伞属 Marasmius 6 2.87
小菇属 Mycena 11 5.26
光柄菇属 Pluteus 5 2.39
红菇属 Russula 17 8.13
栓菌属 Trametes 5 2.39
关闭全屏