发表机构
Beike Language and Intelligence（贝壳语言与智能）
机构由 AI 辅助整理，请以论文原文为准。AI 中文总结
该研究针对记忆增强LLMs的错误传播问题，提出基于SCM的框架，识别两条错误路径并开展多层面评估，发现错误传播规律，路径引导修复可显著降低残留错误。
AI 中文摘要
长期记忆使大型语言模型（LLMs）能够在多轮交互中保存和复用信息，但也可能将局部错误转化为持续风险。现有研究主要评估记忆系统是否正确存储和检索信息，对错误如何在响应、记忆状态和未来交互中传播的理解十分有限。我们提出一种基于结构因果模型（SCM）的框架，用于分析记忆增强LLMs中的跨轮错误传播。我们将用户问题、模型响应和记忆状态建模为动态因果过程，并确定两个错误进入路径：内部记忆更新和外部问题反馈。通过对这些路径进行干预，我们构建了四条反事实轨迹，并量化其下游效应与交互作用。错误影响在四个层面进行评估：记忆保留、自然响应、针对性诊断探测以及概率级错误偏好。实验表明，错误影响通常随交互距离增加而衰减，且记忆更新路径比问题反馈路径产生更持久的效应；潜在错误可能在自然响应中消失后仍存在。传播模式也因记忆类别和记忆机制而异。路径引导修复进一步验证了这种分解：问题修复（Question Repair）可减少27.5%的残留错误，记忆修复（Memory Repair）可减少70.2%，联合修复（Joint Repair）可减少98.3%，几乎消除了残留传播。
英文摘要
Long-term memory enables large language models (LLMs) to preserve and reuse information across interactions, but it can also turn localized errors into persistent risks. Existing work mainly evaluates whether memory systems store and retrieve information correctly, leaving limited understanding of how errors propagate across responses, memory states, and future interactions. We propose a structural causal model (SCM)-based framework for cross-turn error propagation in memory-augmented LLMs. We model user questions, model responses, and memory states as a dynamic causal process, and identify two entry pathways: internal memory updating and external question feedback. By intervening on these pathways, we construct four counterfactual trajectories and quantify their downstream effects and interaction. Error influence is evaluated at four levels: memory retention, natural responses, targeted diagnostic probing, and probability-level error preference. Experiments show that error influence generally decays with interaction distance, while the memory-update pathway contributes more persistent effects than question feedback; latent errors may remain even after disappearing from natural responses. Propagation patterns also vary across memory categories and memory mechanisms. Pathway-guided restoration further validates this decomposition: Question Repair reduces residual error by 27.5%, Memory Repair by 70.2%, and Joint Repair by 98.3%, nearly eliminating residual propagation.
