SGMem：用于长期对话智能体的句子图记忆
SGMem: Sentence Graph Memory for Long-Term Conversational Agents
AI总结：
SGMem通过句子图记忆在分块单元内建模对话关联，结合原始对话与生成记忆，提升长期对话问答的准确率，在LongMemEval和LoCoMo上优于强基线。
摘要：
长期对话智能体需要有效的记忆管理，以处理超出大语言模型（LLMs）上下文窗口的对话历史。现有的基于事实提取或摘要的方法虽然减少了冗余，但在不同粒度的对话和生成的记忆中组织和检索相关信息方面存在困难。我们引入了SGMem（句子图记忆），该方法在分块单元内将对话表示为句子级图，捕获跨轮次、回合和会话级别上下文的关联。通过将检索到的原始对话与生成的记忆（如摘要、事实和见解）相结合，SGMem为LLMs提供连贯且相关的上下文以生成响应。在LongMemEval和LoCoMo上的实验表明，SGMem在长期对话问答中持续提高准确性，并优于强基线方法。
英文摘要：
Long-term conversational agents require effective memory management to handle dialogue histories that exceed the context window of large language models (LLMs). Existing methods based on fact extraction or summarization reduce redundancy but struggle to organize and retrieve relevant information across different granularities of dialogue and generated memory. We introduce SGMem (Sentence Graph Memory), which represents dialogue as sentence-level graphs within chunked units, capturing associations across turn-, round-, and session-level contexts. By combining retrieved raw dialogue with generated memory such as summaries, facts and insights, SGMem supplies LLMs with coherent and relevant context for response generation. Experiments on LongMemEval and LoCoMo show that SGMem consistently improves accuracy and outperforms strong baselines in long-term conversational question answering.
机构、总结、中文摘要均由AI分析而来，可能存在错误，请以论文正文为准。
