{"id":8079,"date":"2026-01-13T06:09:00","date_gmt":"2026-01-12T21:09:00","guid":{"rendered":"https:\/\/devneko.jp\/wordpress\/?p=8079"},"modified":"2026-01-11T15:30:13","modified_gmt":"2026-01-11T06:30:13","slug":"ai-memory%e9%96%a2%e9%80%a3%e3%81%ae%e8%ab%96%e6%96%87%e3%80%81%e3%83%99%e3%83%b3%e3%83%81%e3%83%9e%e3%83%bc%e3%82%af","status":"publish","type":"post","link":"https:\/\/devneko.jp\/wordpress\/?p=8079","title":{"rendered":"AI Memory\u95a2\u9023\u306e\u8ad6\u6587\u3001\u30d9\u30f3\u30c1\u30de\u30fc\u30af"},"content":{"rendered":"\n<p>\u5148\u9031\u306fAI Memory\u95a2\u9023\u306e\u8ad6\u6587\u304c\u591a\u304f\u51fa\u3066\u3044\u305f\u3002\u30d9\u30f3\u30c1\u30de\u30fc\u30af\u3082\u5897\u3048\u3066\u3044\u3066\u91cd\u8981\u304b\u3064\u71b1\u3044\u5206\u91ce\u3002<\/p>\n\n\n\n<ul class=\"wp-block-list\">\n<li><strong>EvolMem: A Cognitive-Driven Benchmark for Multi-Session Dialogue Memory\u00a0<\/strong>[63.8]<br>EvolMem\u306f\u3001\u5927\u898f\u6a21\u8a00\u8a9e\u30e2\u30c7\u30eb(LLM)\u3068\u30a8\u30fc\u30b8\u30a7\u30f3\u30c8\u30b7\u30b9\u30c6\u30e0\u306e\u30de\u30eb\u30c1\u30bb\u30c3\u30b7\u30e7\u30f3\u30e1\u30e2\u30ea\u6a5f\u80fd\u3092\u8a55\u4fa1\u3059\u308b\u305f\u3081\u306e\u65b0\u3057\u3044\u30d9\u30f3\u30c1\u30de\u30fc\u30af\u3067\u3042\u308b\u3002 \u3053\u306e\u30d9\u30f3\u30c1\u30de\u30fc\u30af\u3092\u69cb\u7bc9\u3059\u308b\u305f\u3081\u306b,\u8a71\u984c\u304b\u3089\u59cb\u307e\u308b\u751f\u6210\u3068\u7269\u8a9e\u304b\u3089\u7740\u60f3\u3092\u5f97\u305f\u5909\u63db\u304b\u3089\u306a\u308b\u30cf\u30a4\u30d6\u30ea\u30c3\u30c9\u30c7\u30fc\u30bf\u5408\u6210\u30d5\u30ec\u30fc\u30e0\u30ef\u30fc\u30af\u3092\u63d0\u6848\u3059\u308b\u3002 \u5e83\u7bc4\u306a\u8a55\u4fa1\u306b\u3088\u308a\u3001\u3069\u306eLLM\u3082\u3059\u3079\u3066\u306e\u30e1\u30e2\u30ea\u6b21\u5143\u306b\u304a\u3044\u3066\u4e00\u8cab\u3057\u3066\u4ed6\u3092\u4e0a\u56de\u308b\u3053\u3068\u306f\u306a\u3044\u3053\u3068\u304c\u660e\u3089\u304b\u306b\u306a\u308a\u307e\u3057\u305f\u3002\u00a0<br><a href=\"http:\/\/arxiv.org\/abs\/2601.03543v1\">\u8ad6\u6587<\/a>\u00a0\u00a0<a href=\"https:\/\/fugumt.com\/fugumt\/paper_check\/2601.03543v1\">\u53c2\u8003\u8a33\uff08\u30e1\u30bf\u30c7\u30fc\u30bf\uff09<\/a>\u00a0 \u00a0(Wed, 07 Jan 2026 03:14:42 GMT)<\/li>\n\n\n\n<li>\u30e1\u30e2\u30ea\u6a5f\u80fd\u306e\u305f\u3081\u306e\u30d9\u30f3\u30c1\u30de\u30fc\u30af<\/li>\n\n\n\n<li>\u30ea\u30dd\u30b8\u30c8\u30ea\u306f<a href=\"https:\/\/github.com\/shenye7436\/EvolMem\">GitHub &#8211; shenye7436\/EvolMem<\/a><\/li>\n<\/ul>\n\n\n\n<ul class=\"wp-block-list\">\n<li><strong>Mem-Gallery: Benchmarking Multimodal Long-Term Conversational Memory for MLLM Agents\u00a0<\/strong>[76.8]<br>\u9577\u671f\u8a18\u61b6\u306f\u30de\u30eb\u30c1\u30e2\u30fc\u30c0\u30eb\u5927\u8a00\u8a9e\u30e2\u30c7\u30eb(MLLM)\u30a8\u30fc\u30b8\u30a7\u30f3\u30c8\u306b\u3068\u3063\u3066\u91cd\u8981\u306a\u6a5f\u80fd\u3067\u3042\u308b\u3002 Mem-Gallery\u306fMLLM\u30a8\u30fc\u30b8\u30a7\u30f3\u30c8\u306e\u30de\u30eb\u30c1\u30e2\u30fc\u30c0\u30eb\u9577\u671f\u4f1a\u8a71\u30e1\u30e2\u30ea\u8a55\u4fa1\u306e\u305f\u3081\u306e\u65b0\u3057\u3044\u30d9\u30f3\u30c1\u30de\u30fc\u30af\u3067\u3042\u308b\u3002<br><a href=\"http:\/\/arxiv.org\/abs\/2601.03515v1\">\u8ad6\u6587<\/a>\u00a0\u00a0<a href=\"https:\/\/fugumt.com\/fugumt\/paper_check\/2601.03515v1\">\u53c2\u8003\u8a33\uff08\u30e1\u30bf\u30c7\u30fc\u30bf\uff09<\/a>\u00a0 \u00a0(Wed, 07 Jan 2026 02:03:13 GMT)<\/li>\n\n\n\n<li>\u30de\u30eb\u30c1\u30e2\u30fc\u30c0\u30eb\u6027\u3092\u8003\u616e\u3057\u305f\u30d9\u30f3\u30c1\u30de\u30fc\u30af<\/li>\n\n\n\n<li>\u30ea\u30dd\u30b8\u30c8\u30ea\u306f<a href=\"https:\/\/github.com\/YuanchenBei\/Mem-Gallery\">GitHub &#8211; YuanchenBei\/Mem-Gallery: The source code of Mem-Gallery. Benchmarking Multimodal Long-Term Conversational Memory for MLLM Agents.<\/a><\/li>\n<\/ul>\n\n\n\n<ul class=\"wp-block-list\">\n<li><strong>Agentic Memory: Learning Unified Long-Term and Short-Term Memory Management for Large Language Model Agents\u00a0<\/strong>[57.4]<br>\u5927\u898f\u6a21\u8a00\u8a9e\u30e2\u30c7\u30eb (LLM) \u30a8\u30fc\u30b8\u30a7\u30f3\u30c8\u306f\u3001\u6709\u9650\u30b3\u30f3\u30c6\u30ad\u30b9\u30c8\u30a6\u30a3\u30f3\u30c9\u30a6\u306b\u3088\u308b\u9577\u8ddd\u96e2\u63a8\u8ad6\u306b\u304a\u3044\u3066\u57fa\u672c\u7684\u306a\u5236\u9650\u306b\u76f4\u9762\u3057\u3066\u3044\u308b\u3002 \u65e2\u5b58\u306e\u30e1\u30bd\u30c3\u30c9\u306f\u901a\u5e38\u3001\u9577\u671f\u8a18\u61b6(LTM)\u3068\u77ed\u671f\u8a18\u61b6(STM)\u3092\u72ec\u7acb\u3057\u305f\u30b3\u30f3\u30dd\u30fc\u30cd\u30f3\u30c8\u3068\u3057\u3066\u6271\u3046\u3002 \u672c\u7a3f\u3067\u306f,\u30a8\u30fc\u30b8\u30a7\u30f3\u30c8\u306e\u30dd\u30ea\u30b7\u30fc\u306b LTM \u3068 STM \u7ba1\u7406\u3092\u76f4\u63a5\u7d71\u5408\u3059\u308b\u7d71\u5408\u30d5\u30ec\u30fc\u30e0\u30ef\u30fc\u30af\u3067\u3042\u308b Agentic Memory (AgeMem) \u3092\u63d0\u6848\u3059\u308b\u3002<br><a href=\"http:\/\/arxiv.org\/abs\/2601.01885v1\">\u8ad6\u6587<\/a>\u00a0\u00a0<a href=\"https:\/\/fugumt.com\/fugumt\/paper_check\/2601.01885v1\">\u53c2\u8003\u8a33\uff08\u30e1\u30bf\u30c7\u30fc\u30bf\uff09<\/a>\u00a0 \u00a0(Mon, 05 Jan 2026 08:24:16 GMT)<\/li>\n\n\n\n<li>\u9577\u671f\u30fb\u77ed\u671f\u8a18\u61b6\u3092\u7d71\u4e00\u7684\u306b\u6271\u3046\u30a2\u30d7\u30ed\u30fc\u30c1\u3001\u300cwe propose Agentic Memory (Age- Mem), a unified memory management framework that enables LLM-based agents to jointly control long-term and short-term memory through learn- able, tool-based actions. By integrating memory operations directly into the agent\u2019s policy and training them with a progressive reinforcement learning strategy, AgeMem replaces heuristic memory pipelines with an end-to-end optimized solution. Extensive experiments across diverse long-horizon benchmarks show that AgeMem improves both task performance and memory quality while maintaining efficient context usage.\u300d<\/li>\n<\/ul>\n\n\n\n<ul class=\"wp-block-list\">\n<li><strong>EverMemOS: A Self-Organizing Memory Operating System for Structured Long-Horizon Reasoning\u00a0<\/strong>[42.3]<br>\u5927\u304d\u306a\u8a00\u8a9e\u30e2\u30c7\u30eb(LLM)\u306f\u3001\u9577\u671f\u306e\u5bfe\u8a71\u30a8\u30fc\u30b8\u30a7\u30f3\u30c8\u3068\u3057\u3066\u307e\u3059\u307e\u3059\u30c7\u30d7\u30ed\u30a4\u3055\u308c\u3066\u3044\u308b\u304c\u3001\u305d\u306e\u9650\u3089\u308c\u305f\u30b3\u30f3\u30c6\u30ad\u30b9\u30c8\u30a6\u30a3\u30f3\u30c9\u30a6\u306f\u3001\u62e1\u5f35\u3055\u308c\u305f\u76f8\u4e92\u4f5c\u7528\u3088\u308a\u3082\u30b3\u30d2\u30fc\u30ec\u30f3\u30c8\u306a\u632f\u821e\u3044\u3092\u7dad\u6301\u3059\u308b\u306e\u304c\u56f0\u96e3\u3067\u3042\u308b\u3002 \u672c\u7a3f\u3067\u306f,EverMemOS\u306b\u3064\u3044\u3066\u7d39\u4ecb\u3059\u308b\u3002EverMemOS\u306f,\u8a08\u7b97\u30e1\u30e2\u30ea\u306b\u30a8\u30df\u30e5\u30ec\u30fc\u30c8\u3055\u308c\u305f\u30e9\u30a4\u30d5\u30b5\u30a4\u30af\u30eb\u3092\u5b9f\u88c5\u3057\u305f\u81ea\u5df1\u7d44\u7e54\u578b\u30e1\u30e2\u30ea\u30aa\u30da\u30ec\u30fc\u30c6\u30a3\u30f3\u30b0\u30b7\u30b9\u30c6\u30e0\u3067\u3042\u308b\u3002 EverMemOS\u306f\u3001\u30e1\u30e2\u30ea\u62e1\u5f35\u63a8\u8ad6\u30bf\u30b9\u30af\u3067\u6700\u5148\u7aef\u306e\u30d1\u30d5\u30a9\u30fc\u30de\u30f3\u30b9\u3092\u9054\u6210\u3059\u308b\u3002<br><a href=\"http:\/\/arxiv.org\/abs\/2601.02163v1\">\u8ad6\u6587<\/a>\u00a0\u00a0<a href=\"https:\/\/fugumt.com\/fugumt\/paper_check\/2601.02163v1\">\u53c2\u8003\u8a33\uff08\u30e1\u30bf\u30c7\u30fc\u30bf\uff09<\/a>\u00a0 \u00a0(Mon, 05 Jan 2026 14:39:43 GMT)<\/li>\n\n\n\n<li>\u300cWe introduce EverMemOS, a self-organizing memory operating system that implements an engram- inspired lifecycle for computational memory. Episodic Trace Formation converts dialogue streams into MemCells that capture episodic traces, atomic facts, and time-bounded Foresight signals. Semantic Consolidation organizes MemCells into thematic MemScenes, distilling stable semantic structures and updating user profiles. Reconstructive Recollection per- forms MemScene-guided agentic retrieval to compose the necessary and sufficient context for downstream reasoning. Experiments on LoCoMo and LongMemEval show that EverMemOS achieves state-of-the-art performance on memory-augmented reasoning tasks.\u300d\u3068\u306e\u3053\u3068<\/li>\n\n\n\n<li>\u30ea\u30dd\u30b8\u30c8\u30ea\u306f<a href=\"https:\/\/github.com\/EverMind-AI\/EverMemOS\">GitHub &#8211; EverMind-AI\/EverMemOS: EverMemOS is an open-source, enterprise-grade intelligent memory system. Our mission is to build AI memory that never forgets, making every conversation built on previous understanding.<\/a><\/li>\n<\/ul>\n\n\n\n<ul class=\"wp-block-list\">\n<li><strong>Controllable Memory Usage: Balancing Anchoring and Innovation in Long-Term Human-Agent Interaction\u00a0<\/strong>[35.2]<br>\u30a8\u30fc\u30b8\u30a7\u30f3\u30c8\u306e\u30e1\u30e2\u30ea\u4f9d\u5b58\u3092\u660e\u793a\u7684\u304b\u3064\u30e6\u30fc\u30b6\u5236\u5fa1\u53ef\u80fd\u306a\u6b21\u5143\u3068\u3057\u3066\u30e2\u30c7\u30eb\u5316\u3067\u304d\u308b\u3053\u3068\u3092\u793a\u3059\u3002 Steerable Memory Agent, SteeM\u3092\u63d0\u6848\u3059\u308b\u3002<br><a href=\"http:\/\/arxiv.org\/abs\/2601.05107v1\">\u8ad6\u6587<\/a>\u00a0\u00a0<a href=\"https:\/\/fugumt.com\/fugumt\/paper_check\/2601.05107v1\">\u53c2\u8003\u8a33\uff08\u30e1\u30bf\u30c7\u30fc\u30bf\uff09<\/a>\u00a0 \u00a0(Thu, 08 Jan 2026 16:54:30 GMT)<\/li>\n\n\n\n<li>\u300cWe then propose Steerable Memory Agent, SteeM, a framework that allows users to dynamically regulate memory reliance, ranging from a fresh- start mode that promotes innovation to a high- fidelity mode that closely follows interaction history.\u300d\u3068Memory\u306e\u5229\u7528\u5ea6\u3092\u5236\u5fa1\u3059\u308b\u30a2\u30a4\u30c7\u30a2<\/li>\n<\/ul>\n\n\n\n<ul class=\"wp-block-list\">\n<li><strong>SimpleMem: Efficient Lifelong Memory for LLM Agents\u00a0<\/strong>[73.7]<br>\u30bb\u30de\u30f3\u30c6\u30a3\u30c3\u30af\u30ed\u30b9\u30ec\u30b9\u5727\u7e2e\u306b\u57fa\u3065\u304f\u52b9\u7387\u7684\u306a\u30e1\u30e2\u30ea\u30d5\u30ec\u30fc\u30e0\u30ef\u30fc\u30afSimpleMem\u3092\u7d39\u4ecb\u3059\u308b\u3002 \u672c\u7a3f\u3067\u306f,\u60c5\u5831\u5bc6\u5ea6\u3068\u30c8\u30fc\u30af\u30f3\u5229\u7528\u91cf\u306e\u6700\u5927\u5316\u3092\u76ee\u7684\u3068\u3057\u305f3\u6bb5\u968e\u30d1\u30a4\u30d7\u30e9\u30a4\u30f3\u3092\u63d0\u6848\u3059\u308b\u3002 \u30d9\u30f3\u30c1\u30de\u30fc\u30af\u30c7\u30fc\u30bf\u30bb\u30c3\u30c8\u3092\u7528\u3044\u305f\u5b9f\u9a13\u306b\u3088\u308a,\u63d0\u6848\u624b\u6cd5\u306f\u7cbe\u5ea6,\u691c\u7d22\u52b9\u7387,\u63a8\u8ad6\u30b3\u30b9\u30c8\u306b\u304a\u3044\u3066,\u30d9\u30fc\u30b9\u30e9\u30a4\u30f3\u30a2\u30d7\u30ed\u30fc\u30c1\u3092\u4e00\u8cab\u3057\u3066\u4e0a\u56de\u3063\u3066\u3044\u308b\u3053\u3068\u304c\u308f\u304b\u3063\u305f\u3002<br><a href=\"http:\/\/arxiv.org\/abs\/2601.02553v1\">\u8ad6\u6587<\/a>\u00a0\u00a0<a href=\"https:\/\/fugumt.com\/fugumt\/paper_check\/2601.02553v1\">\u53c2\u8003\u8a33\uff08\u30e1\u30bf\u30c7\u30fc\u30bf\uff09<\/a>\u00a0 \u00a0(Mon, 05 Jan 2026 21:02:49 GMT)<\/li>\n\n\n\n<li>\u300cSimpleMem mitigates context inflation through three stages. (1) Semantic Structured Compression filters redundant interaction content and reformulates raw dialogue into compact, context-independent memory units.  (2) Recursive Consolidation incrementally organizes related memory units into higher-level abstract representations, reducing redundancy in long-term memory. (3) Adaptive Query-Aware Retrieval dynamically adjusts retrieval scope based on query complexity, enabling efficient context construction under constrained token budgets.\u300d\u3068\u3044\u3046\u30a2\u30d7\u30ed\u30fc\u30c1\u3002\u52b9\u679c\u306f\u5927\u304d\u305d\u3046\u3067\u306f\u3042\u308b\u3082\u306e\u306e\u3001\u3053\u308c\u3092\u3082\u3063\u3066\u300cSemantic Lossless Compression\u300d\u3068\u3044\u3063\u3066\u3088\u3044\u306e\u3060\u308d\u3046\u304b\u3068\u3044\u3046\u306e\u306f\u82e5\u5e72\u7591\u554f\u3002<\/li>\n\n\n\n<li>\u30ea\u30dd\u30b8\u30c8\u30ea\u306f<a href=\"https:\/\/github.com\/aiming-lab\/SimpleMem\">GitHub &#8211; aiming-lab\/SimpleMem: SimpleMem: Efficient Lifelong Memory for LLM Agents<\/a><\/li>\n<\/ul>\n","protected":false},"excerpt":{"rendered":"<p>\u5148\u9031\u306fAI Memory\u95a2\u9023\u306e\u8ad6\u6587\u304c\u591a\u304f\u51fa\u3066\u3044\u305f\u3002\u30d9\u30f3\u30c1\u30de\u30fc\u30af\u3082\u5897\u3048\u3066\u3044\u3066\u91cd\u8981\u304b\u3064\u71b1\u3044\u5206\u91ce\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":[244,387],"class_list":["post-8079","post","type-post","status-publish","format-standard","hentry","category-arxiv","tag-memory","tag-survey"],"_links":{"self":[{"href":"https:\/\/devneko.jp\/wordpress\/index.php?rest_route=\/wp\/v2\/posts\/8079","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=8079"}],"version-history":[{"count":1,"href":"https:\/\/devneko.jp\/wordpress\/index.php?rest_route=\/wp\/v2\/posts\/8079\/revisions"}],"predecessor-version":[{"id":8080,"href":"https:\/\/devneko.jp\/wordpress\/index.php?rest_route=\/wp\/v2\/posts\/8079\/revisions\/8080"}],"wp:attachment":[{"href":"https:\/\/devneko.jp\/wordpress\/index.php?rest_route=%2Fwp%2Fv2%2Fmedia&parent=8079"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/devneko.jp\/wordpress\/index.php?rest_route=%2Fwp%2Fv2%2Fcategories&post=8079"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/devneko.jp\/wordpress\/index.php?rest_route=%2Fwp%2Fv2%2Ftags&post=8079"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}