{"id":7534,"date":"2025-10-07T06:09:00","date_gmt":"2025-10-06T21:09:00","guid":{"rendered":"https:\/\/devneko.jp\/wordpress\/?p=7534"},"modified":"2025-10-04T17:17:25","modified_gmt":"2025-10-04T08:17:25","slug":"mem-%ce%b1-learning-memory-construction-via-reinforcement-learning","status":"publish","type":"post","link":"https:\/\/devneko.jp\/wordpress\/?p=7534","title":{"rendered":"Mem-\u03b1: Learning Memory Construction via Reinforcement Learning\u00a0"},"content":{"rendered":"\n<ul class=\"wp-block-list\">\n<li><strong>Mem-\u03b1: Learning Memory Construction via Reinforcement Learning&nbsp;<\/strong>[20.9]<br>\u5927\u304d\u306a\u8a00\u8a9e\u30e2\u30c7\u30eb(LLM)\u30a8\u30fc\u30b8\u30a7\u30f3\u30c8\u306f\u3001\u9650\u3089\u308c\u305f\u30b3\u30f3\u30c6\u30ad\u30b9\u30c8\u30a6\u30a3\u30f3\u30c9\u30a6\u306b\u3088\u3063\u3066\u5236\u7d04\u3055\u308c\u308b\u3002 \u73fe\u5728\u306e\u30e1\u30e2\u30ea\u62e1\u5f35\u30a8\u30fc\u30b8\u30a7\u30f3\u30c8\u306f\u3001\u30e1\u30e2\u30ea\u66f4\u65b0\u306e\u305f\u3081\u306e\u4e8b\u524d\u306b\u5b9a\u7fa9\u3055\u308c\u305f\u547d\u4ee4\u3068\u30c4\u30fc\u30eb\u306b\u4f9d\u5b58\u3057\u3066\u3044\u308b\u3002 Mem-alpha\u306f\u3001\u30a8\u30fc\u30b8\u30a7\u30f3\u30c8\u306b\u8907\u96d1\u306a\u30e1\u30e2\u30ea\u30b7\u30b9\u30c6\u30e0\u3092\u52b9\u679c\u7684\u306b\u7ba1\u7406\u3059\u308b\u3088\u3046\u306b\u8a13\u7df4\u3059\u308b\u5f37\u5316\u5b66\u7fd2\u30d5\u30ec\u30fc\u30e0\u30ef\u30fc\u30af\u3067\u3042\u308b\u3002<br><a href=\"http:\/\/arxiv.org\/abs\/2509.25911v1\">\u8ad6\u6587<\/a>&nbsp;&nbsp;<a href=\"https:\/\/fugumt.com\/fugumt\/paper_check\/2509.25911v1\">\u53c2\u8003\u8a33\uff08\u30e1\u30bf\u30c7\u30fc\u30bf\uff09<\/a>&nbsp; &nbsp;(Tue, 30 Sep 2025 08:02:34 GMT)<\/li>\n\n\n\n<li>\u30b7\u30b9\u30c6\u30e0\u30d7\u30ed\u30f3\u30d7\u30c8\u7b49\u3067\u51e6\u7406\u3092\u884c\u3046\u30e1\u30e2\u30ea\u7ba1\u7406\u30a8\u30fc\u30b8\u30a7\u30f3\u30c8\u3067\u306f\u9650\u754c\u304c\u3042\u308b\u305f\u3081\u30e1\u30e2\u30ea\u7ba1\u7406\u6226\u7565\u3092\u5b66\u3076\u3088\u3046\u5f37\u5316\u5b66\u7fd2\u3092\u6d3b\u7528\u3059\u308b\u30a2\u30d7\u30ed\u30fc\u30c1\u3092\u63d0\u6848\u300cwe propose Mem-\u03b1, a reinforcement learning framework that trains agents to effectively manage complex memory systems through interaction and feedback.&nbsp;\u300d<\/li>\n\n\n\n<li>\u300cEmpirical evaluation demonstrates that Mem-\u03b1 achieves significant improvements over existing memory-augmented agent baselines across diverse benchmarks. Most remarkably, despite being trained exclusively on instances with a maximum length of 30k tokens, our agents exhibit robust generalization to sequences exceeding 400k tokens, over 13\u00d7 the training length.\u300d\u3068\u3044\u3046\u306e\u3082\u8208\u5473\u6df1\u3044\u3002<\/li>\n\n\n\n<li>\u30ea\u30dd\u30b8\u30c8\u30ea\u306f<a href=\"https:\/\/github.com\/wangyu-ustc\/Mem-alpha\">GitHub &#8211; wangyu-ustc\/Mem-alpha: Learning Memory Construction via Reinforcement Learning<\/a><\/li>\n<\/ul>\n","protected":false},"excerpt":{"rendered":"","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-7534","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\/7534","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=7534"}],"version-history":[{"count":2,"href":"https:\/\/devneko.jp\/wordpress\/index.php?rest_route=\/wp\/v2\/posts\/7534\/revisions"}],"predecessor-version":[{"id":7537,"href":"https:\/\/devneko.jp\/wordpress\/index.php?rest_route=\/wp\/v2\/posts\/7534\/revisions\/7537"}],"wp:attachment":[{"href":"https:\/\/devneko.jp\/wordpress\/index.php?rest_route=%2Fwp%2Fv2%2Fmedia&parent=7534"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/devneko.jp\/wordpress\/index.php?rest_route=%2Fwp%2Fv2%2Fcategories&post=7534"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/devneko.jp\/wordpress\/index.php?rest_route=%2Fwp%2Fv2%2Ftags&post=7534"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}