{"id":7083,"date":"2025-07-14T04:55:00","date_gmt":"2025-07-13T19:55:00","guid":{"rendered":"https:\/\/devneko.jp\/wordpress\/?p=7083"},"modified":"2025-07-12T15:12:27","modified_gmt":"2025-07-12T06:12:27","slug":"memos-a-memory-os-for-ai-system-mirix-multi-agent-memory-system-for-llm-based-agents","status":"publish","type":"post","link":"https:\/\/devneko.jp\/wordpress\/?p=7083","title":{"rendered":"MemOS: A Memory OS for AI System, MIRIX: Multi-Agent Memory System for LLM-Based Agents"},"content":{"rendered":"\n<p>RAG\u3067\u306f\u53b3\u3057\u3044\u554f\u984c\u3092\u6271\u3046\u305f\u3081\u306eMemory\u95a2\u9023\u306e\u7814\u7a76\u304c\u3068\u3066\u3082\u76db\u3093\u3002<\/p>\n\n\n\n<ul class=\"wp-block-list\">\n<li><strong>MemOS: A Memory OS for AI System\u00a0<\/strong>[115.3]<br>\u5927\u898f\u6a21\u8a00\u8a9e\u30e2\u30c7\u30eb(LLM)\u306f\u3001\u4eba\u5de5\u77e5\u80fd(AGI)\u306b\u3068\u3063\u3066\u4e0d\u53ef\u6b20\u306a\u57fa\u76e4\u3068\u306a\u3063\u3066\u3044\u308b\u3002 \u65e2\u5b58\u306e\u30e2\u30c7\u30eb\u306f\u3001\u4e3b\u306b\u9759\u7684\u30d1\u30e9\u30e1\u30fc\u30bf\u3068\u77ed\u547d\u306a\u30b3\u30f3\u30c6\u30ad\u30b9\u30c8\u72b6\u614b\u306b\u4f9d\u5b58\u3057\u3066\u304a\u308a\u3001\u30e6\u30fc\u30b6\u306e\u597d\u307f\u3092\u8ffd\u8de1\u3057\u305f\u308a\u3001\u9577\u3044\u671f\u9593\u306b\u308f\u305f\u3063\u3066\u77e5\u8b58\u3092\u66f4\u65b0\u3059\u308b\u80fd\u529b\u3092\u5236\u9650\u3059\u308b\u3002 MemOS\u306f\u30e1\u30e2\u30ea\u3092\u7ba1\u7406\u53ef\u80fd\u306a\u30b7\u30b9\u30c6\u30e0\u30ea\u30bd\u30fc\u30b9\u3068\u3057\u3066\u6271\u3046\u30e1\u30e2\u30ea\u30aa\u30da\u30ec\u30fc\u30c6\u30a3\u30f3\u30b0\u30b7\u30b9\u30c6\u30e0\u3067\u3042\u308b\u3002<br><a href=\"http:\/\/arxiv.org\/abs\/2507.03724v1\">\u8ad6\u6587<\/a>\u00a0\u00a0<a href=\"https:\/\/fugumt.com\/fugumt\/paper_check\/2507.03724v1\">\u53c2\u8003\u8a33\uff08\u30e1\u30bf\u30c7\u30fc\u30bf\uff09<\/a>\u00a0 \u00a0(Fri, 04 Jul 2025 17:21:46 GMT)<\/li>\n\n\n\n<li><a href=\"https:\/\/devneko.jp\/wordpress\/?p=6856\">MemOS: An Operating System for Memory-Augmented Generation (MAG) in Large Language Models \u2013 arXiv\u6700\u65b0\u8ad6\u6587\u306e\u7d39\u4ecb<\/a>\u304b\u3089\u306e\u30a2\u30c3\u30d7\u30c7\u30fc\u30c8\u3001Agentic\u306a\u30a2\u30d7\u30ed\u30fc\u30c1\u306eLLM\u7528\u30e1\u30e2\u30ea\u3002\u6642\u7cfb\u5217\u6027\u306a\u3069\u901a\u5e38\u306eRAG\u3067\u306f\u7c21\u5358\u3067\u306f\u306a\u3044\u90e8\u5206\u306e\u6027\u80fd\u5411\u4e0a\u304c\u5927\u304d\u3044\u3002\uff08\u304c\u3001\u300cTo ensure architectural parity, all methods are implemented over the same LLM backbone (GPT-4o-mini)\u300d\u3068\u30d9\u30fc\u30b9\u30e2\u30c7\u30eb\u304cGPT-4o mini\u3067\u826f\u3044\u306e\u304b\u306f\u82e5\u5e72\u8b0e\u3067\u306f\u3042\u308b\uff09<\/li>\n\n\n\n<li>\u30ea\u30dd\u30b8\u30c8\u30ea\u306f<a href=\"https:\/\/github.com\/MemTensor\/MemOS\">GitHub &#8211; MemTensor\/MemOS: MemOS (Preview) | Intelligence Begins with Memory<\/a><\/li>\n<\/ul>\n\n\n\n<ul class=\"wp-block-list\">\n<li><strong>MIRIX: Multi-Agent Memory System for LLM-Based Agents\u00a0<\/strong>[7.1]<br>MIRIX\u306f\u8a00\u8a9e\u30e2\u30c7\u30eb\u306e\u305f\u3081\u306e\u30e2\u30b8\u30e5\u30fc\u30eb\u578b\u30de\u30eb\u30c1\u30a8\u30fc\u30b8\u30a7\u30f3\u30c8\u30e1\u30e2\u30ea\u30b7\u30b9\u30c6\u30e0\u3067\u3042\u308b\u3002 MIRIX\u306f\u3001\u30ea\u30c3\u30c1\u306a\u8996\u899a\u7684\u304a\u3088\u3073\u30de\u30eb\u30c1\u30e2\u30fc\u30c0\u30eb\u4f53\u9a13\u3092\u53d7\u3051\u5165\u308c\u308b\u305f\u3081\u306b\u30c6\u30ad\u30b9\u30c8\u3092\u8d85\u8d8a\u3059\u308b\u3002 MIRIX\u306f\u30e1\u30e2\u30ea\u62e1\u5f35LDM\u30a8\u30fc\u30b8\u30a7\u30f3\u30c8\u306e\u65b0\u305f\u306a\u30d1\u30d5\u30a9\u30fc\u30de\u30f3\u30b9\u6a19\u6e96\u3092\u8a2d\u5b9a\u3057\u3066\u3044\u308b\u3002<br><a href=\"http:\/\/arxiv.org\/abs\/2507.07957v1\">\u8ad6\u6587<\/a>\u00a0\u00a0<a href=\"https:\/\/fugumt.com\/fugumt\/paper_check\/2507.07957v1\">\u53c2\u8003\u8a33\uff08\u30e1\u30bf\u30c7\u30fc\u30bf\uff09<\/a>\u00a0 \u00a0(Thu, 10 Jul 2025 17:40:11 GMT)<\/li>\n\n\n\n<li>\u3053\u3061\u3089\u3082Agentic\u306a\u30a2\u30d7\u30ed\u30fc\u30c1\u306e\u30e1\u30e2\u30ea\u7ba1\u7406\u30d5\u30ec\u30fc\u30e0\u30ef\u30fc\u30af\u3002\u30d9\u30fc\u30b9\u30e2\u30c7\u30eb\u304c\u7570\u306a\u308b\u305f\u3081MemOS\u3068\u76f4\u63a5\u6bd4\u8f03\u304c\u56f0\u96e3\u3060\u304c\u3001\u4ed6\u30b7\u30b9\u30c6\u30e0\u3068\u6bd4\u3079\u9ad8\u3044\u6027\u80fd\u3092\u4e3b\u5f35\u3002<\/li>\n\n\n\n<li>\u30ea\u30dd\u30b8\u30c8\u30ea\u306f<a href=\"https:\/\/github.com\/Mirix-AI\/MIRIX\">GitHub &#8211; Mirix-AI\/MIRIX: Mirix is a multi-agent personal assistant designed to track on-screen activities and answer user questions intelligently. By capturing real-time visual data and consolidating it into structured memories, Mirix transforms raw inputs into a rich knowledge base that adapts to your digital experiences.<\/a><\/li>\n<\/ul>\n\n\n\n<ul class=\"wp-block-list\">\n<li><strong>Evaluating Memory in LLM Agents via Incremental Multi-Turn Interactions\u00a0<\/strong>[19.5]<br>\u30e1\u30e2\u30ea\u6a5f\u69cb\u3092\u6301\u3064\u30a8\u30fc\u30b8\u30a7\u30f3\u30c8\u3092\u30e1\u30e2\u30ea\u30a8\u30fc\u30b8\u30a7\u30f3\u30c8\u3068\u547c\u3076\u3002 \u672c\u7a3f\u3067\u306f,\u30e1\u30e2\u30ea\u30a8\u30fc\u30b8\u30a7\u30f3\u30c8\u306b\u4e0d\u53ef\u6b20\u306a4\u3064\u306e\u30b3\u30a2\u80fd\u529b,\u3059\u306a\u308f\u3061,\u6b63\u78ba\u306a\u691c\u7d22,\u30c6\u30b9\u30c8\u6642\u9593\u5b66\u7fd2,\u9577\u8ddd\u96e2\u7406\u89e3,\u30b3\u30f3\u30d5\u30ea\u30af\u30c8\u89e3\u6c7a\u306e4\u3064\u3092\u540c\u5b9a\u3059\u308b\u3002 \u65e2\u5b58\u306e\u30c7\u30fc\u30bf\u30bb\u30c3\u30c8\u306f\u3001\u9650\u3089\u308c\u305f\u30b3\u30f3\u30c6\u30ad\u30b9\u30c8\u9577\u306b\u4f9d\u5b58\u3059\u308b\u304b\u3001\u66f8\u7c4d\u30d9\u30fc\u30b9\u306eQA\u306e\u3088\u3046\u306a\u9759\u7684\u3067\u9577\u3044\u30b3\u30f3\u30c6\u30ad\u30b9\u30c8\u8a2d\u5b9a\u7528\u306b\u8abf\u6574\u3055\u308c\u3066\u3044\u308b\u3002 \u65e2\u5b58\u306e\u30d9\u30f3\u30c1\u30de\u30fc\u30af\u3067\u306f4\u3064\u306e\u80fd\u529b\u3092\u3059\u3079\u3066\u30ab\u30d0\u30fc\u3057\u3066\u3044\u306a\u3044\u305f\u3081\u3001\u30e1\u30e2\u30ea\u30a8\u30fc\u30b8\u30a7\u30f3\u30c8\u7528\u306b\u7279\u5225\u306b\u8a2d\u8a08\u3055\u308c\u305f\u65b0\u3057\u3044\u30d9\u30f3\u30c1\u30de\u30fc\u30af\u3067\u3042\u308bMemoryAgentBench\u3092\u7d39\u4ecb\u3057\u307e\u3059\u3002<br><a href=\"http:\/\/arxiv.org\/abs\/2507.05257v1\">\u8ad6\u6587<\/a>\u00a0\u00a0<a href=\"https:\/\/fugumt.com\/fugumt\/paper_check\/2507.05257v1\">\u53c2\u8003\u8a33\uff08\u30e1\u30bf\u30c7\u30fc\u30bf\uff09<\/a>\u00a0 \u00a0(Mon, 07 Jul 2025 17:59:54 GMT)<\/li>\n\n\n\n<li>\u3053\u3061\u3089\u306fMemory\u3092\u6301\u3064\u30a8\u30fc\u30b8\u30a7\u30f3\u30c8\u306e\u305f\u3081\u306e\u30d9\u30f3\u30c1\u30de\u30fc\u30af\u306e\u63d0\u6848<\/li>\n\n\n\n<li>\u300cwe identify four core competencies essential for memory agents: accurate retrieval, test-time learning, long-range understanding, and conflict resolution.\u300d\u3068\u306e\u3053\u3068\u3002<\/li>\n\n\n\n<li>\u7d50\u679c\u306b\u3042\u308b\u300cWhile Mem0 has demonstrated relatively strong performance on conversational tasks such as LOCOMO\u2014where information density is comparatively low\u2014it tends to perform poorly on benchmarks containing dense informational content, including RULER and \u221e-Bench. For tasks emphasizing Time-to-Live (TTL) and Least Recently Used (LRU) retrieval, these limitations are often even more pronounced.\u300d\u3068\u3044\u3046\u6307\u6458\u306f\u8208\u5473\u6df1\u304f\u3001\u30c9\u30e1\u30a4\u30f3\u3092\u9078\u3070\u306a\u3044\u6c4e\u7528\u7684\u306a\u69cb\u9020\u3092\u4f5c\u308b\u306e\u306f\u5927\u5909\u305d\u3046\u3068\u3044\u3046\u5370\u8c61\u3002<\/li>\n\n\n\n<li>\u30ea\u30dd\u30b8\u30c8\u30ea\u306f<a href=\"https:\/\/huggingface.co\/datasets\/ai-hyz\/MemoryAgentBench\">ai-hyz\/MemoryAgentBench \u00b7 Datasets at Hugging Face<\/a>\u3001<a href=\"https:\/\/github.com\/HUST-AI-HYZ\/MemoryAgentBench\">GitHub &#8211; HUST-AI-HYZ\/MemoryAgentBench: Open source code for Paper: Evaluating Memory in LLM Agents via Incremental Multi-Turn Interactions<\/a><\/li>\n<\/ul>\n","protected":false},"excerpt":{"rendered":"<p>RAG\u3067\u306f\u53b3\u3057\u3044\u554f\u984c\u3092\u6271\u3046\u305f\u3081\u306eMemory\u95a2\u9023\u306e\u7814\u7a76\u304c\u3068\u3066\u3082\u76db\u3093\u3002<\/p>\n","protected":false},"author":1,"featured_media":0,"comment_status":"open","ping_status":"open","sticky":false,"template":"","format":"standard","meta":{"footnotes":""},"categories":[2],"tags":[244],"class_list":["post-7083","post","type-post","status-publish","format-standard","hentry","category-arxiv","tag-memory"],"_links":{"self":[{"href":"https:\/\/devneko.jp\/wordpress\/index.php?rest_route=\/wp\/v2\/posts\/7083","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/devneko.jp\/wordpress\/index.php?rest_route=\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/devneko.jp\/wordpress\/index.php?rest_route=\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/devneko.jp\/wordpress\/index.php?rest_route=\/wp\/v2\/users\/1"}],"replies":[{"embeddable":true,"href":"https:\/\/devneko.jp\/wordpress\/index.php?rest_route=%2Fwp%2Fv2%2Fcomments&post=7083"}],"version-history":[{"count":1,"href":"https:\/\/devneko.jp\/wordpress\/index.php?rest_route=\/wp\/v2\/posts\/7083\/revisions"}],"predecessor-version":[{"id":7084,"href":"https:\/\/devneko.jp\/wordpress\/index.php?rest_route=\/wp\/v2\/posts\/7083\/revisions\/7084"}],"wp:attachment":[{"href":"https:\/\/devneko.jp\/wordpress\/index.php?rest_route=%2Fwp%2Fv2%2Fmedia&parent=7083"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/devneko.jp\/wordpress\/index.php?rest_route=%2Fwp%2Fv2%2Fcategories&post=7083"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/devneko.jp\/wordpress\/index.php?rest_route=%2Fwp%2Fv2%2Ftags&post=7083"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}