让本地AI代理像人一样记住对话，会遗忘还会自动整理记忆。
SuperLocalMemory V3.3: The Living Brain -- Biologically-Inspired Forgetting, Cognitive Quantization, and Multi-Channel Retrieval for Zero-LLM Agent Memory Systems
- 用生物启发的遗忘机制和多通道检索，模拟人类记忆过程。
- 在零LLM模式下实现70.4%准确率，多跳任务提升23.8个百分点。
- 完全本地运行、无需云端，适合隐私敏感或离线场景使用。
AI编程代理面临悖论：拥有海量参数知识却记不住一小时前的对话。现有记忆系统将文本存于向量库中，仅支持单通道检索，依赖云端LLM进行核心操作，且未实现人类记忆的有效认知过程。我们提出SuperLocalMemory V3.3（‘活脑’），一个以本地优先的代理记忆系统，实现了完整的认知记忆分类体系与数学化生命周期动态。基于V3.2（arXiv:2603.14588）的信息几何基础，提出五项贡献：(1) Fisher-Rao量化感知距离（FRQAD）——在高斯统计流形上定义的新度量，在偏好高保真嵌入方面实现100%精度（对比余弦为85.6%），无先例；(2) Ebbinghaus自适应遗忘与生命周期感知量化——首次在本地代理记忆中引入数学化遗忘曲线并耦合渐进嵌入压缩，判别力提升6.7倍；(3) 7通道认知检索，涵盖语义、关键词、实体图、时间、扩散激活、巩固与霍普菲尔德关联通道，在零LLM模式下达到LoCoMo测试70.4%准确率；(4) 通过软提示实现长期隐式记忆参数化；(5) 零摩擦自动认知流水线，实现完整记忆生命周期自动化。在LoCoMo上，V3.3在模式A（零LLM）下取得70.4%表现，多跳任务+23.8个百分点，对抗任务+12.7个百分点。V3.2在模式A为74.8%，模式C为87.7%；4.4个百分点差距为有意架构权衡。SLM V3.3开源，采用Elastic License 2.0，纯CPU运行，月下载量超5000次。
原文摘要 · Abstract (English)
AI coding agents operate in a paradox: they possess vast parametric knowledge yet cannot remember a conversation from an hour ago. Existing memory systems store text in vector databases with single-channel retrieval, require cloud LLMs for core operations, and implement none of the cognitive processes that make human memory effective. We present SuperLocalMemory V3.3 ("The Living Brain"), a local-first agent memory system implementing the full cognitive memory taxonomy with mathematical lifecycle dynamics. Building on the information-geometric foundations of V3.2 (arXiv:2603.14588), we introduce five contributions: (1) Fisher-Rao Quantization-Aware Distance (FRQAD) -- a new metric on the Gaussian statistical manifold achieving 100% precision at preferring high-fidelity embeddings over quantized ones (vs 85.6% for cosine), with zero prior art; (2) Ebbinghaus Adaptive Forgetting with lifecycle-aware quantization -- the first mathematical forgetting curve in local agent memory coupled to progressive embedding compression, achieving 6.7x discriminative power; (3) 7-channel cognitive retrieval spanning semantic, keyword, entity graph, temporal, spreading activation, consolidation, and Hopfield associative channels, achieving 70.4% on LoCoMo in zero-LLM Mode A; (4) memory parameterization implementing Long-Term Implicit memory via soft prompts; (5) zero-friction auto-cognitive pipeline automating the complete memory lifecycle. On LoCoMo, V3.3 achieves 70.4% in Mode A (zero-LLM), with +23.8pp on multi-hop and +12.7pp on adversarial. V3.2 achieved 74.8% Mode A and 87.7% Mode C; the 4.4pp gap reflects a deliberate architectural trade-off. SLM V3.3 is open source under the Elastic License 2.0, runs entirely on CPU, with over 5,000 monthly downloads.
Thank you to arXiv for use of its open access interoperability. PaperDance 不是 arXiv 官方产品；中文卡片由大模型生成，请以原文为准。