Tag
AAMAS 2027 论文提出将 LLM 智能体持久记忆的治理分为「准入」与「呈现」两个决策，并提出两种免重训练的推理时设计（FC 与 PS），在外部基准上分别降低 6.7 与 8.8 个百分点的失败率，同时显著减少跨领域信息泄露，但均加剧个性化失败，表明准入选择可带来任务特定的安全收益而有益记忆的保留仍未解决。
MemGuard introduces a system that persists verifier signals as metadata to govern memory in LLM agents, improving reliability and performance across multiple benchmarks like SWE-Bench and WebArena.
The author launches an open-source repository and field guide for governed memory in autonomous and agentic systems, covering what becomes memory, uncertainty, influence on behavior, and correction/forgetting. Contributions and discussions are welcome.
An analysis of the current state of AI memory systems, arguing that the focus has shifted from storing more data to defining how memory should behave—covering governance, observability, lifecycle management, and interoperability.
MemoryOps AI is an open-source memory governance layer for AI assistants that handles memory lifecycle with policies, expiration, auditing, and deletion guarantees. The author seeks technical feedback from developers building AI agents and RAG systems.
GateMem is a benchmark for evaluating memory governance in multi-principal shared-memory agents, covering utility, access control, and forgetting across medical, office, education, and household domains. Current methods fail to balance all three, showing that reliable shared institutional deployment remains elusive.
A Memory Curator Agent pattern separates memory governance from worker agents, improving the quality and relevance of durable memory by having a dedicated agent evaluate and route memory events into appropriate scopes.