Tag
论文提出 HeRE，一个通过 LLM 引导的知识图谱对齐去噪构建的希伯来语关系抽取基准与 150 万条银标数据集，并提出可推广至其他低资源语言的可复现流水线。研究发现多语言模型在少样本场景优于特定语言模型，而 Gemma-4-31B-it (F1=0.915) 已接近 GPT-5.4 (0.921) 的表现。
A viral post shares a 15-minute setup combining Claude Code with tools like graphify (folder-to-knowledge-graph, 126k GitHub stars and YC-backed), googleworkspace CLI, Playwright MCP, and jev-guard to build a self-checking personal agent with memory, scheduling, and safety guards.
文章论述检索型智能体不应将所有数据压平为文本块，而应暴露 KB、DB、Graph、Skills、Memory 等多类型数据界面，由检索代理按需选择最小充分集合，并已在 OpenDCAI/DataMind 中实现。
The paper proposes OpCredit, an operation-aware policy-learning framework for agentic knowledge graph question answering that combines Branch Outcome Credit and Frontier Transition Credit to assign credit to individual operations instead of whole trajectories, showing consistent gains on ComplexWebQuestions, WebQSP, and GrailQA.
FinGREAT 评估开放权重 LLM 在金融图引导检索与关系分类任务上的表现，利用 11 节点 11 边的复式记账图谱分析 1300 份 SEC 10-K 文件，并引入包含 483 段人工标注的关系分类基准，最强开源与闭源模型准确率分别为 84.1% 和 92.8%。
SurrealDB 创始人 Tobie 宣布推出基于其多模型数据库的 Agent Memory，可将对话和文档转化为带溯源与时序历史的结构化实体、关系和事实，并在 BEAM 基准上达到 77.2% 准确率。
Oracle AI Agent Memory 26.8 introduces graph-aware retrieval, letting agents store memories as linked records with typed relationships (supersedes, supports, contradicts) and expand search results beyond flat semantic similarity. The release also adds image-aware memory, schema upgrades, and enterprise controls.
yggy 是一款自我构建的 AI 知识图谱产品，可将杂乱文档与连接系统自动转化为带来源追溯的知识图谱，无需设计模式或搭建管道，支持团队应用、API 与 MCP 智能体接入，并托管于欧盟。
该论文介绍了团队参加 KnowledgeGraphEval 2026 子任务2（阿拉伯语关系抽取）的系统：结合基于编码器的方法与微调 Yehia、Qwen3.5 的 LLM 系统，最佳集成系统在子任务2中以 94.06 的 micro-F1 获得第一名，凸显了约束预测与多种建模方法互补的重要性。
Explains how Jev can act as a semantic gate before Graphiti for agent memory systems, distinguishing lasting preferences from one-time instructions or sensitive data based on app-specific retention rules. The post clarifies that Jev is only needed when custom product-level memory policies go beyond Graphiti's built-in extraction and provenance capabilities.
KGLite 发布：一个嵌入式、支持 Cypher 查询的知识图谱工具，面向 Python 与 Rust，专为 AI 应用和 LLM 代理设计，可进程内运行、从 DataFrame 批量加载，并内置 MCP 服务器，采用 MIT 开源协议。
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。
Jev-Mem:一篇来自UT Dallas的论文提出用TypeSafe AI的Jev结构化输出模型替代LLM来处理智能体记忆系统中的大量小型决策（如记忆分类、关联判断、检索停止条件），仅在最终生成答案时才调用LLM，从而显著降低token开销。
Atlassian and OpenAI are expanding their partnership, bringing GPT-6 family frontier models across Atlassian's platform and Rovo to power AI agents grounded in the Teamwork Graph enterprise context layer. OpenAI will also rely on Jira internally, and both companies are exploring deeper integrations for assigning work to AI agents and measuring their impact with Atlassian's DX.
作者开源了一套「AI 赋能学习/教学/科研」的工具仓库，用 AI 提炼知识点依赖图、规划学习路线、生成讲义、播客、错题本和学习报告，以解决教材中「递归学习」难题。
Multiple teams independently built AI agents that each maintained isolated memory over the same accounts, causing conflicting actions at renewal time. The fix was moving account history to a shared live store—DevRev's Computer synced with HubSpot, Gmail, and Slack—accessed by all agents via MCP.
The article explores how splitting tasks across conversation turns reduces agent accuracy and proposes a memory system using context-specific retrieval types to maintain accuracy, with a hands-on example using the open-source Graphiti engine.
VeriChron AI is an open-source bitemporal compliance and audit platform that reconstructs historical states using knowledge graphs and AI agents to prevent future information leakage in audits.
The paper introduces CodeGraph, a large-scale open-taxonomy knowledge graph for source code, built using a specialized LLM and grounded in Wikidata, covering millions of files and entities.
This paper presents WaterBERT, a domain-adapted BERT model for water treatment literature mining, enhancing semantic representation and enabling large-scale structured information extraction and knowledge graph construction.