Tag
提出CogMem认知记忆架构，基于PEC²F图谱将对话记忆从被动检索转为主动重构，在LoCoMo和LongMemEval上取得多跳、时序和知识更新任务的良好表现。
Hindsight 是 Vectorize 开源的 Agent 记忆系统，超越简单的聊天记录检索，通过 Retain/Recall/Reflect 三个动作让 Agent 从过往经验中形成新判断，支持 Claude Code、Codex、Cursor 等并内置 MCP Server。
CharacterMemory is an open-source memory system for AI characters inspired by neuroscience, retrieving information based on emotion, recency and connections, with world/emotion tracking, routine and calendar management, a WebUI, and easy integration into existing code.
The author released Memoria 1.0.0, a local-first, model-agnostic memory system for LLM applications that runs without a GPU or cloud API key. Benchmarked on an Intel Celeron N4020 (3.7 GiB RAM), it achieved 89.8% Recall@1 and 99.6% Recall@50 on LongMemEval-S using parallel FAISS, BM25, graph, phrase, attribute, and temporal retrieval with multi-signal ranking, and is installable via PyPI.
ChipMEM introduces a verification-grounded memory layer for EDA agents, enhancing RTL design performance by distilling reusable skills from verified tool interactions and using Bayesian statistical guidance for transfer to unseen tasks.
LightMem-Ego, a system for AI glasses to remember daily life from visual/audio streams, has received a major upgrade with new features and a real-time demo.
The paper introduces EmbodiedMemory-Bench, a benchmark for evaluating memory in long-horizon embodied tasks, and presents Embodied-Memorizer, an external memory system with an 8B policy model to enhance agent performance.
The paper proposes SpeakerMem-R1, a speaker-centered dual-track memory system for multi-party dialogue, addressing bottlenecks in message attribution and relational understanding. It achieves state-of-the-art results on benchmarks like EverMemBench and LoCoMo.
A local, file-based long-term memory layer for personal AI that stores raw conversation logs with timestamps, enabling time-first search and maintaining full context without summarization.
The article explains how agent memory systems can leak data due to post-filter tenant scoping in vector search and recommends scoping at write time to prevent cross-tenant data exposure.
This study compares memory systems to full conversation history in AI agents over simulated days, showing 23-62x fewer context tokens with similar or better recall on personal facts, but memory loses on numerical data and specific details like identifiers.
Thread is an AI journal and second brain that captures thoughts, ideas, and recordings, automatically connects related memories, and helps users turn scattered ideas into actionable insights.
The article proposes a Markdown-based memory system for AI agents that uses simple file storage and editing rules to outperform complex memory runtimes, ensuring accurate and sourced fact retrieval.
A factory worker built an autonomous AI entity using Gemma 4 31B on Ollama, with self-identity, memory, and a blog, demonstrating personalized AI interaction through an open-source framework.
Garry Tan has released new evaluations for gbrain, an open-source retrieval layer for AI agents, showcasing state-of-the-art performance in memory reading and writing without LLM-in-loop retrieval.
An open-source project enables turning any LLM into a 3D character that talks, moves, and reacts in the browser with memory of emotional states and relationships, fully customizable and free to use.
The team is phasing out the Membrane system in favor of the more powerful Meron, whose core breakthrough involves introducing the 'meaning atoms' concept. By storing raw information and multiple representations, it enables flexible and powerful AI memory.
The author built a memory layer for AI agents called fireweed, which uses deterministic code to verify claims against evidence, preventing hallucinations and enabling auditable, provable memory erasure.
The paper introduces the Weighted Memory Tree, a hierarchical memory system for LLM agents that dynamically retains important information, improving task accuracy by 9.97% and reducing prompt token usage by 32.8%.
This paper introduces StateMemBench, a benchmark for evaluating state tracking in LLM agent memory systems, and proposes StateMem, a method that improves current-state accuracy in long interactions.