[论文] An Interpretable Memory Decision Controller for LLM Agents Based on Th...

论文概要

研究领域: NLP 作者: Yiming Zhang, Jinghong Zhang, Haoran Zhao, Yiren Ma, Chunlei Zhao 发布时间: 2026-09-18 arXiv: 2609.22043

中文摘要

大语言模型的记忆系统大多关注高效检索，而“检索到的记忆是否应被信任”这一决策受到的关注相对较少。当记忆库包含冲突立场时，标准检索增强生成（RAG）会盲目注入记忆并放大幻觉：在易受记忆注入影响的模型中，冲突记忆下的 RAG 幻觉率明显高于无记忆基线。受前额叶皮层记忆信号机制启发，我们提出记忆决策层（MDL）——一个位于检索与生成阶段之间的零参数记忆决策控制器。其核心是三信号互补编码器，通过基于 QR 的正交子空间投影将相关性、可靠性和任务风险融合，并结合元工作记忆信号，形成可解释的决策表征，量化检索记忆的可信度。在此基础上，MDL 显式解耦置信度与一致性，并引入风险反转和显式弃权。在主流大语言模型和多个开源数据集上的评估表明，MDL 在一般场景下将冲突记忆下的幻觉率降低约 56.04%，在高风险场景下趋近于零幻觉。该控制器是完全白盒的：纯几何操作、无需训练参数，每次决策仅增加约 0.14 毫秒——约为前置嵌入检索步骤的 50 倍之快，比 LLM 自评估调用快四到五个数量级。

原文摘要

Memory systems for large language models have focused predominantly on efficient retrieval, whereas the decision of whether retrieved memories should be trusted has received comparatively little attention. When the memory store contains conflicting positions, standard retrieval-augmented generation (RAG) blindly injects memories and amplifies hallucinations: in models susceptible to memory injection, the RAG hallucination rate under conflicting memories is markedly higher than that of a memory-free baseline. Inspired by memory signaling mechanisms in the prefrontal cortex, we propose the Memory Decision Layer (MDL), a zero-parameter memory decision controller situated between the retrieval and generation stages. Its core is a three-signal complementary encoder that fuses relevance, reliabi...

#论文 #arXiv #NLP #小凯