ES-Mem: Event Segmentation-Based Memory for Long-Term Dialogue Agents
College of Computer Science, Chongqing University(重庆大学计算机科学学院);Tsinghua University(清华大学);Hong Kong Generative AI Research & Development Center, HKUST(香港科技大学生成式人工智能研究与开发中心);Aerospace Information Research Institute, Chinese Academy of Sciences(中国科学院航空航天信息研究所);School of Computer Science and Technology, Tiangong University(天津理工大学计算机科学与技术学院)
Memory is critical for dialogue agents to maintain coherence and enable continuous adaptation in long-term interactions. While existing memory mechanisms offer basic storage and retrieval capabilities, they are hindered by two primary limitations: (1) rigid memory granularity often disrupts semantic integrity, resulting in fragmented and incoherent memory units; (2) prevalent flat retrieval paradigms rely solely on surface-level semantic similarity, neglecting the structural cues of discourse required to navigate and locate specific episodic contexts. To mitigate these limitations, drawing inspiration from Event Segmentation Theory, we propose ES-Mem, a framework incorporating two core components: (1) a dynamic event segmentation module that partitions long-term interactions into semantically coherent events with distinct boundaries; (2) a hierarchical memory architecture that constructs multi-layered memories and leverages boundary semantics to anchor specific episodic memory for precise context localization. Evaluations on two memory benchmarks demonstrate that ES-Mem yields consistent performance gains over baseline methods. Furthermore, the proposed event segmentation module exhibits robust applicability on dialogue segmentation datasets.
Turn Segmentation into Utterances for Arabic Spontaneous Dialogues and Instance Messages
Institute of Statistical Studies and Research (ISSR), Cairo University(开罗大学统计研究与研究所);Faculty of Computers and Information, Cairo University(开罗大学计算机与信息学院)
Recursively Summarizing Enables Long-Term Dialogue Memory in Large Language Models
the Department of Computer Science and Engineering, Hong Kong University of Science and Technology(香港科技大学计算机科学与工程系);the institute of information engineering, Chinese Academy of Sciences(中国科学院信息工程研究所);the College of Computer, National University of Defense Technology(国防科技大学计算机学院);the University of Sydney(悉尼大学)
Dialogue Session Segmentation by Embedding-Enhanced TextTiling
Peking University(北京大学);Key Laboratory of High Confidence Software Technologies (Peking University), MoE(高可信软件技术教育部重点实验室(北京大学));Natural Language Processing Department, Baidu Inc.(百度公司自然语言处理部);School of Computer, Central China Normal University(华中师范大学计算机学院)
Dialogue Act Segmentation for Vietnamese Human-Human Conversational Texts
Thainguyen University (TNU)(太原大学(TNU));University of Information and Communication Technology(信息与通信技术学院);Vietnam National University, Hanoi (VNU)(河内越南国家大学(VNU));University of Engineering and Technology(工程技术学院)
解析口语对话带来了独特的困难,包括不流利和类似于广告的单元之间的无标记边界。以前的工作已经表明,韵律可以帮助解析不流利的语音(Tran et al. 2018),但假设解析器的输入已经被分割成类似韵律的单元(SU),这在现有的语音应用中是不正确的。我们研究韵律如何影响一个分析器,接收整个对话轮作为输入(一个回合为基础的模型),而不是黄金标准的预分割SU(一个SU为基础的模型)。在英语Switchboard语料库上的实验中,我们发现,当单独使用成绩单,话轮为基础的模型有麻烦分割SU,导致更差的分析性能比SU为基础的模型。然而,韵律可以有效地取代黄金标准SU边界:韵律,回合为基础的模型表现以及SU为基础的模型(90.79与90.65 F1得分,分别),尽管执行两个任务(SU分割和解析),而不是一个(解析单独)。分析表明,音高和强度特征对于该语料库是最重要的,因为它们允许模型正确区分SU边界和语音不流利-这是模型在其他方面难以做到的区别。
Memory Consolidation for Contextual Spoken Language Understanding with Dialogue Logistic Inference
Institute of Automation, CAS(中国科学院自动化研究所);University of Chinese Academy of Sciences(中国科学院大学);CAS Center for Excellence in Brain Science and Intelligence Technology(中国科学院脑科学与智能技术卓越创新中心)
VSTAR: A Video-grounded Dialogue Dataset for Situated Semantic Understanding with Scene and Topic Transitions
Wangxuan Institute of Computer Technology, Peking University(北京大学王选计算机研究所);Beijing Institute for General Artificial Intelligence(北京通用人工智能研究院);National Key Laboratory of General Artificial Intelligence(通用人工智能全国重点实验室)
MOOM: Maintenance, Organization and Optimization of Memory in Ultra-Long Role-Playing Dialogues
Beijing University of Posts and Telecommunications(北京邮电大学);SenseTime(商汤科技);Beijing Key Laboratory of Network System and Network Culture(北京网络系统与网络文化重点实验室)
Efficient Dialogue State Tracking by Selectively Overwriting Memory
Clova AI, NAVER Corp.(NAVER公司Clova AI)
最近的作品在对话状态跟踪(DST)集中在一个开放的基于词汇的设置,以解决可扩展性和泛化问题的预定义的基于本体的方法。然而,它们是低效的,因为它们从头开始预测每一个回合的对话状态。在这里,我们认为对话状态作为一个显式的固定大小的内存,并提出了一个选择性的存储机制,更有效的DST。该机制由两个步骤组成:(1)预测每个存储器槽上的状态操作,以及(2)用新的值来存储存储器,其中只有少数是根据预测的状态操作生成的。该方法将DST分解为两个子任务,并引导解码器只关注其中一个任务,从而减轻了解码器的负担。这提高了培训和DST性能的有效性。我们的SOM-DST(Selectively Overwriting Memory for Dialogue State Tracking)模型在基于开放词汇的DST设置中实现了最先进的联合目标准确率,MultiWOZ 2.0为51.72%,MultiWOZ 2.1为53.01%。此外,我们分析了当前和地面真值给定情况之间的准确性差距,并建议改善状态操作预测以提高DST性能是一个有前途的方向。
RealMem: Benchmarking LLMs in Real-World Memory-Driven Interaction
Xidian University(西安电子科技大学);Zhejiang University(浙江大学);Peking University(北京大学);Shanghai Jiao Tong University(上海交通大学);Renmin University of China(中国人民大学);Sun Yat-sen University(中山大学);University of the Chinese Academy of Sciences(中国科学院大学)
HingeMem: Boundary Guided Long-Term Memory with Query Adaptive Retrieval for Scalable Dialogues
College of Design and Innovation, Tongji University(同济大学设计创意学院);Shanghai Research Institute for Intelligent Autonomous Systems, Tongji University(同济大学上海自主智能无人系统科学中心)