Conversational search (CS) needs a holistic understanding of conversational inputs to retrieve relevant passages. In this paper, we demonstrate the existence of a retrieval shortcut in CS, which causes models to retrieve passages solely relying on partial history while disregarding the latest question. With in-depth analysis, we first show that naively trained dense retrievers heavily exploit the shortcut and hence perform poorly when asked to answer history-independent questions. To build more robust models against shortcut dependency, we explore various hard negative mining strategies. Experimental results show that training with the model-based hard negatives effectively mitigates the dependency on the shortcut, significantly improving dense retrievers on recent CS benchmarks. In particular, our retriever outperforms the previous state-of-the-art model by 11.0 in Recall@10 on QReCC.
ConvMix: A Mixed-Criteria Data Augmentation Framework for Conversational Dense Retrieval
Fengran Mo 1（摩峰ran 1）；Jinghan Zhang 2（张 Jinghan 2）；Yuchen Hui 1（慧 Yuchen 1）；Jia Ao Sun 1（Sun Jia Ao 1）；Zhichao Xu 3（徐 Zhichao 3）；Zhan Su 1（苏 Zhan 1）；Jian-Yun Nie 1（Nie Jian-Yun 1）
UniRetriever: Multi-task Candidates Selection for Various Context-Adaptive Conversational Retrieval
The Chinese University of Hong Kong（香港中文大学）；Nankai University（南开大学）；Harbin Institute of Technology, Shenzhen（哈尔滨工业大学（深圳））；Huawei Noah’s Ark Lab（华为诺亚方舟实验室）
Learning to Relate to Previous Turns in Conversational Search
University of Montreal（蒙特利尔大学）；Institute for AI Industry Research, Tsinghua University（清华大学人工智能产业研究院）；Gaoling School of Artificial Intelligence, Renmin University of China（中国人民大学高瓴人工智能学院）；Department of Computer Science, Tsinghua University（清华大学计算机科学系）
ChatRetriever: Adapting Large Language Models for Generalized and Robust Conversational Dense Retrieval
Gaoling School of Artificial Intelligence, Renmin University of China（中国人民大学高瓴人工智能学院）；Université de Montréal（蒙特利尔大学）；Beijing Academy of Artificial Intelligence（北京人工智能研究院）；Waseda University（早稻田大学）
ConvSDG: Session Data Generation for Conversational Search
Université de Montréal（蒙特利尔大学）；RALI（RALI（机器翻译与计算语言学研究实验室））；Renmin University of China（中国人民大学）；University of Massachusetts Amherst（马萨诸塞大学阿默斯特分校）；Beijing Jiaotong University（北京交通大学）
University of Montreal（蒙特利尔大学）；University of Massachusetts Amherst（马萨诸塞大学阿默斯特分校）；Renmin University of China（中国人民大学）；Beihang University（北京航空航天大学）；University of Copenhagen（哥本哈根大学）；Beijing Jiaotong University（北京交通大学）
Building Open-Retrieval Conversational Question Answering Systems by Generating Synthetic Data and Decontextualizing User Questions
Department of Informatics, Athens University of Economics and Business（雅典经济与商业大学信息学院）；Omilia - Conversational Intelligence（Omilia-对话智能）；Archimedes, Athena Research Center（雅典研究中心Archimedes）；NCSR Demokritos（德摩斯蒂斯国家研究中心）
Aligning Query Representation with Rewritten Query and Relevance Judgments in Conversational Search
Université de Montréal（蒙特利尔大学）；University of Massachusetts Amherst（马萨诸塞大学阿默斯特分校）；Renmin University of China（中国人民大学）；University of Copenhagen（哥本哈根大学）；Beijing Jiaotong University（北京交通大学）
