TEM: Temporal State Tracking Memory for Large Language Model Agents
We present Temporal State Tracking Memory (TEM), a training-free framework for tracking these changes.
ProofPaper ↗
Key points
- An agent's memory must remain useful as information changes across conversations.
- A new fact may replace an earlier value without making that value irrelevant to a later question.
- TEM stores facts as versioned records linked to their sources.
- A further evaluation on 1,540 LoCoMo questions tested hybrid memory under a fixed evidence budget.
Sources (1)
- [1]TEM: Temporal State Tracking Memory for Large Language Model AgentsarXiv (AI, ML, NLP, CV, robotics, multi-agent) · Oct 15, 01:54 PM
We present Temporal State Tracking Memory (TEM), a training-free framework for tracking these changes.
An agent's memory must remain useful as information changes across conversations.
Extractive summary: sentences quoted from the sources.
Before this
- Oct 15, 2026Loud and Clear: Making Visual Evidence Count in Multimodal Document RAG
- Oct 15, 2026Beyond Retrieval Failure: Understanding Why LLMs Miss Evidence in Multi-Hop RAG
- Oct 14, 2026Generate presentations with Amazon Bedrock AgentCore
- Oct 14, 2026jaikydota/BullFlow: 推量牛 BullFlow — 开源企业级 RAG 解决方案:从企业知识库、营销内容自动化生产到全域内容分发 | Open-source enterprise RAG solution: knowledge base → automated marketing content →
- Oct 14, 20262430816465/campus-knowledge-agent: 校园知识服务 Agent 智能体平台:基于 FastAPI、Qwen3 与 RAG,贯通公开资料解析、时间范围检索、证据溯源及人工审核。
- Oct 13, 2026anisayari/sparky: 🔒 Your private AI assistant, 100% local on Windows: chat, vision, documents (RAG), web search, images, videos with sound, voice, code projec