大语言模型中的记忆：机制、评估与演化
Memory in Large Language Models: Mechanisms, Evaluation and Evolution
AI总结：
本文提出大语言模型记忆的统一操作性定义、四部分分类法和记忆四元组，并通过分层评估、时序治理、泄漏审计及DMM Gov更新遗忘机制，构建可复现、可比较且可治理的研究与部署坐标系统。
摘要：
在统一的操作性定义下，我们将大语言模型（LLM）记忆定义为在预训练、微调或推理过程中写入的持久状态，该状态之后可被寻址，并稳定地影响输出。我们提出一个四部分分类法（参数化、上下文、外部、程序性/情景性）以及一个记忆四元组（位置、持久性、写入/访问路径、可控性）。我们通过“写入 -> 读取 -> 抑制/更新”链条将机制、评估和治理联系起来。为避免在异构设置下产生失真的比较，我们采用三设置协议（仅参数化、离线检索、在线检索），在相同数据和时序上将能力与信息可用性解耦。在此基础上，我们构建了分层评估：参数化（闭卷回忆、编辑差分、记忆/隐私）、上下文（位置曲线和中段下降）、外部（答案正确性 vs 片段归因/忠实性），以及程序性/情景性（跨会话一致性和时序回放，E MARS+）。该框架整合了时序治理和泄漏审计（新鲜度命中、过时答案、拒绝切片），并通过评分者间一致性加带多重比较校正的配对检验进行不确定性报告。针对更新与遗忘，我们提出 DMM Gov：协调 DAPT/TAPT、PEFT、模型编辑（ROME、MEND、MEMIT、SERAC）和 RAG，形成可审计的闭环，覆盖准入阈值、上线、监控、回滚和变更审计，并包含及时性、冲突处理和长时程一致性的规范。最后，我们给出四个可检验命题：最小可识别性；最小评估卡；带可验证遗忘的因果约束编辑；以及小窗口回放检索何时优于超长上下文阅读。这为研究和部署提供了一个可复现、可比较且可治理的坐标系统。
英文摘要：
Under a unified operational definition, we define LLM memory as a persistent state written during pretraining, finetuning, or inference that can later be addressed and that stably influences outputs. We propose a four-part taxonomy (parametric, contextual, external, procedural/episodic) and a memory quadruple (location, persistence, write/access path, controllability). We link mechanism, evaluation, and governance via the chain write -> read -> inhibit/update. To avoid distorted comparisons across heterogeneous setups, we adopt a three-setting protocol (parametric only, offline retrieval, online retrieval) that decouples capability from information availability on the same data and timeline. On this basis we build a layered evaluation: parametric (closed-book recall, edit differential, memorization/privacy), contextual (position curves and the mid-sequence drop), external (answer correctness vs snippet attribution/faithfulness), and procedural/episodic (cross-session consistency and timeline replay, E MARS+). The framework integrates temporal governance and leakage auditing (freshness hits, outdated answers, refusal slices) and uncertainty reporting via inter-rater agreement plus paired tests with multiple-comparison correction. For updating and forgetting, we present DMM Gov: coordinating DAPT/TAPT, PEFT, model editing (ROME, MEND, MEMIT, SERAC), and RAG to form an auditable loop covering admission thresholds, rollout, monitoring, rollback, and change audits, with specs for timeliness, conflict handling, and long-horizon consistency. Finally, we give four testable propositions: minimum identifiability; a minimal evaluation card; causally constrained editing with verifiable forgetting; and when retrieval with small-window replay outperforms ultra-long-context reading. This yields a reproducible, comparable, and governable coordinate system for research and deployment.
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。
