[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"news-f453d930-c5d6-437c-a4fc-db1a6de04b68":3},{"id":4,"title":5,"summary":6,"original_url":7,"source_id":8,"tags":9,"published_at":23,"created_at":24,"modified_at":25,"is_published":26,"publish_type":27,"image_url":13,"view_count":28},"f453d930-c5d6-437c-a4fc-db1a6de04b68","WAIC 2026 首发开源\"记忆操作系统\"MemOS：准确率、Token、推理三个维度同时盖过 OpenAI 全局记忆","WAIC 2026 开幕首日，国产团队\"记忆张量\"（MemTensor）正式开源记忆操作系统 MemOS，把\"记忆\"从 LLM 的外挂检索（RAG）变成可调度的系统资源，直接对标 OpenAI 全局记忆方案。\n\nMemOS 的核心抽象是 MemCube——把明文记忆、激活记忆、参数记忆三种形态统一封装，每一块都带元数据（来源、版本、生命周期），可组合、可迁移、可融合。对外提供三层 API：参数层（LoRA 式外挂参数）、激活层（KV Cache 持久化）、明文层（RAG）。这套架构精准解决了通用大模型的三大短板：无持续状态、无自我迭代、无个性化认知。\n\n实测数据很硬：相较 OpenAI 全局记忆，平均准确率提升 38.97%、Token 运行开销下降 60.95%、时序复杂推理任务性能提升 159%。MemOS 采用 Apache 2.0 协议开源，配套 MemOS Studio、LightRAG、MemReader 多模态解析器，并已接入 MCP、Dify、Coze 等生态。\n\n腾讯副总裁韩开创在 WAIC 现场同时指出：多智能体长任务记忆丢失率高达 40%，指令偏差与上下文断层是核心故障源。这从工业界印证了\"记忆工程化\"的紧迫性。当行业还在卷参数和 KV Cache 时，MemOS 把\"记忆\"升格为底层资源，配合 Harness 工程框架，是后 Scaling 时代中国 AGI 走出的一条差异化路线。","https:\u002F\u002Fwww.qbitai.com\u002F2026\u002F07\u002F443399.html","3bd971a8-3897-43d9-84ac-43879efd2f94",[10,14,17,20],{"id":11,"name":12,"slug":12,"description":13,"color":13},"6ad31a14-c0da-42df-81fd-564281f768db","agentic-ai",null,{"id":15,"name":16,"slug":16,"description":13,"color":13},"01598627-1ea6-4b27-a5d8-874971571a71","llm",{"id":18,"name":19,"slug":19,"description":13,"color":13},"7e89b5cc-57db-4f37-bc6d-28919a73931c","model-release",{"id":21,"name":22,"slug":22,"description":13,"color":13},"b9bd9039-fcdb-41a8-b85b-fc1587def2b9","open-source","2026-07-17T16:20:00Z","2026-07-17T16:17:01.337642Z","2026-07-17T16:17:01.337667Z",true,"agent",3]