<?xml version="1.0" encoding="UTF-8"?>
<modsCollection xmlns="http://www.loc.gov/mods/v3">
<mods ID="ding-etal-2023-ji">
    <titleInfo>
        <title>基于推理链的多跳问答对抗攻击和对抗增强训练方法(Reasoning Chain Based Adversarial Attack and Adversarial Augmentation Training for Multi-hop Question Answering)</title>
    </titleInfo>
    <name type="personal">
        <namePart type="given">Jiayu</namePart>
        <namePart type="family">Ding</namePart>
            <roleTerm authority="marcrelator" type="text">author</roleTerm>
        </role>
    </name>
    <name type="personal">
        <namePart type="given">Siyuan</namePart>
        <namePart type="family">Wang</namePart>
            <roleTerm authority="marcrelator" type="text">author</roleTerm>
        </role>
    </name>
    <name type="personal">
        <namePart type="given">Zhongyu</namePart>
        <namePart type="family">Wei</namePart>
            <roleTerm authority="marcrelator" type="text">author</roleTerm>
        </role>
    </name>
    <name type="personal">
        <namePart type="given">Qin</namePart>
        <namePart type="family">Chen</namePart>
            <roleTerm authority="marcrelator" type="text">author</roleTerm>
        </role>
    </name>
    <name type="personal">
        <namePart type="given">Xuanjing</namePart>
        <namePart type="family">Huang</namePart>
            <roleTerm authority="marcrelator" type="text">author</roleTerm>
        </role>
    </name>
    <originInfo>
        <dateIssued>2023-08</dateIssued>
    </originInfo>
    <typeOfResource>text</typeOfResource>
    <language>
        <languageTerm type="text">Chinese</languageTerm>
        <languageTerm type="code" authority="iso639-2b">chi</languageTerm>
    </language>
    <relatedItem type="host">
        <titleInfo>
            <title>Proceedings of the 22nd Chinese National Conference on Computational Linguistics</title>
        </titleInfo>
        <name type="personal">
            <namePart type="given">Maosong</namePart>
            <namePart type="family">Sun</namePart>
                <roleTerm authority="marcrelator" type="text">editor</roleTerm>
            </role>
        </name>
        <name type="personal">
            <namePart type="given">Bing</namePart>
            <namePart type="family">Qin</namePart>
                <roleTerm authority="marcrelator" type="text">editor</roleTerm>
            </role>
        </name>
        <name type="personal">
            <namePart type="given">Xipeng</namePart>
            <namePart type="family">Qiu</namePart>
                <roleTerm authority="marcrelator" type="text">editor</roleTerm>
            </role>
        </name>
        <name type="personal">
            <namePart type="given">Jing</namePart>
            <namePart type="family">Jiang</namePart>
                <roleTerm authority="marcrelator" type="text">editor</roleTerm>
            </role>
        </name>
        <name type="personal">
            <namePart type="given">Xianpei</namePart>
            <namePart type="family">Han</namePart>
                <roleTerm authority="marcrelator" type="text">editor</roleTerm>
            </role>
        </name>
        <originInfo>
            <publisher>Chinese Information Processing Society of China</publisher>
            <place>
                <placeTerm type="text">Harbin, China</placeTerm>
            </place>
        </originInfo>
        <genre authority="marcgt">conference publication</genre>
    </relatedItem>
    <abstract>“本文提出了一种基于多跳推理链的对抗攻击方法,通过向输入文本中加入对抗性的攻击文本,并测试问答模型在干扰数据下生成答案的准确性,以检测问答模型真正执行多跳推理的能力和可解释性。该方法首先从输入文本中抽取从问题实体到答案实体的推理链,并基于推理链的特征把多跳问题分为了不同的推理类型,提出了一个模型来自动化实现问题拆解和推理类型预测,然后根据推理类型对原问题进行修改来构造攻击干扰句。实验对多个多跳问答模型进行了对抗攻击测试,所有模型的性能都显著下降,验证了该攻击方法的有效性以及目前问答模型存在的不足;向原训练集中加入对抗样本进行增强训练后,模型性能均有所回升,证明了本对抗增强训练方法可以提升模型的鲁棒性。”</abstract>
    <identifier type="citekey">ding-etal-2023-ji</identifier>
    <location>
        <url>https://aclanthology.org/2023.ccl-1.1</url>
    </location>
        <date>2023-08</date>
        <extent unit="page">
            <start>1</start>
            <end>16</end>
        </extent>
    </part>
</mods>
</modsCollection>

%0 Conference Proceedings
%T 基于推理链的多跳问答对抗攻击和对抗增强训练方法(Reasoning Chain Based Adversarial Attack and Adversarial Augmentation Training for Multi-hop Question Answering)
%A Ding, Jiayu
%A Wang, Siyuan
%A Wei, Zhongyu
%A Chen, Qin
%A Huang, Xuanjing
%Y Sun, Maosong
%Y Qin, Bing
%Y Qiu, Xipeng
%Y Jiang, Jing
%Y Han, Xianpei
%S Proceedings of the 22nd Chinese National Conference on Computational Linguistics
%D 2023
%8 August
%I Chinese Information Processing Society of China
%C Harbin, China
%G Chinese
%F ding-etal-2023-ji
%X “本文提出了一种基于多跳推理链的对抗攻击方法,通过向输入文本中加入对抗性的攻击文本,并测试问答模型在干扰数据下生成答案的准确性,以检测问答模型真正执行多跳推理的能力和可解释性。该方法首先从输入文本中抽取从问题实体到答案实体的推理链,并基于推理链的特征把多跳问题分为了不同的推理类型,提出了一个模型来自动化实现问题拆解和推理类型预测,然后根据推理类型对原问题进行修改来构造攻击干扰句。实验对多个多跳问答模型进行了对抗攻击测试,所有模型的性能都显著下降,验证了该攻击方法的有效性以及目前问答模型存在的不足;向原训练集中加入对抗样本进行增强训练后,模型性能均有所回升,证明了本对抗增强训练方法可以提升模型的鲁棒性。”
%U https://aclanthology.org/2023.ccl-1.1
%P 1-16

Markdown (Informal)

[基于推理链的多跳问答对抗攻击和对抗增强训练方法(Reasoning Chain Based Adversarial Attack and Adversarial Augmentation Training for Multi-hop Question Answering)](https://aclanthology.org/2023.ccl-1.1) (Ding et al., CCL 2023)

• 基于推理链的多跳问答对抗攻击和对抗增强训练方法(Reasoning Chain Based Adversarial Attack and Adversarial Augmentation Training for Multi-hop Question Answering) (Ding et al., CCL 2023)

ACL

• Jiayu Ding, Siyuan Wang, Zhongyu Wei, Qin Chen, and Xuanjing Huang. 2023. 基于推理链的多跳问答对抗攻击和对抗增强训练方法(Reasoning Chain Based Adversarial Attack and Adversarial Augmentation Training for Multi-hop Question Answering) . In Proceedings of the 22nd Chinese National Conference on Computational Linguistics , pages 1–16, Harbin, China. Chinese Information Processing Society of China.