{"id":917,"date":"2026-02-05T01:30:13","date_gmt":"2026-02-04T16:30:13","guid":{"rendered":"https:\/\/itexplore.org\/jp\/columns\/ai-trends-brain-networks-llm-knowledge-information-processing\/"},"modified":"2026-02-05T01:30:13","modified_gmt":"2026-02-04T16:30:13","slug":"ai-trends-brain-networks-llm-knowledge-information-processing","status":"publish","type":"post","link":"https:\/\/itexplore.org\/jp\/columns\/ai-trends-brain-networks-llm-knowledge-information-processing\/","title":{"rendered":"AI\u6280\u8853\u306e\u6700\u65b0\u52d5\u5411\uff1a\u8133\u30cd\u30c3\u30c8\u30ef\u30fc\u30af\u3001LLM\u306e\u77e5\u8b58\u3001\u60c5\u5831\u51e6\u7406"},"content":{"rendered":"<p>\u672c\u65e5\u306e\u6ce8\u76eeAI\u30fb\u30c6\u30c3\u30af\u30cb\u30e5\u30fc\u30b9\u3092\u3001\u5c02\u9580\u7684\u306a\u5206\u6790\u3068\u5171\u306b\u304a\u5c4a\u3051\u3057\u307e\u3059\u3002<\/p>\n<div class=\"wp-block-vk-blocks-alert vk_alert alert alert-warning has-alert-icon\">\n<div class=\"vk_alert_icon\">\n<div class=\"vk_alert_icon_icon\"><i class=\"fa-solid fa-triangle-exclamation\" aria-hidden=\"true\"><\/i><\/div>\n<div class=\"vk_alert_icon_text\"><span>Warning<\/span><\/div>\n<\/div>\n<div class=\"vk_alert_content\">\n<p>\u3053\u306e\u8a18\u4e8b\u306fAI\u306b\u3088\u3063\u3066\u81ea\u52d5\u751f\u6210\u30fb\u5206\u6790\u3055\u308c\u305f\u3082\u306e\u3067\u3059\u3002AI\u306e\u6027\u8cea\u4e0a\u3001\u4e8b\u5b9f\u8aa4\u8a8d\u304c\u542b\u307e\u308c\u308b\u53ef\u80fd\u6027\u304c\u3042\u308b\u305f\u3081\u3001\u91cd\u8981\u306a\u5224\u65ad\u3092\u4e0b\u3059\u969b\u306f\u5fc5\u305a\u30ea\u30f3\u30af\u5148\u306e\u4e00\u6b21\u30bd\u30fc\u30b9\u3092\u3054\u78ba\u8a8d\u304f\u3060\u3055\u3044\u3002<\/p>\n<\/div>\n<\/div>\n<div class=\"wp-block-group\" style=\"margin-top:40px;margin-bottom:40px\">\n<h2 class=\"wp-block-heading\">\u8133\u30cd\u30c3\u30c8\u30ef\u30fc\u30af\u8868\u73fe\u306e\u305f\u3081\u306e\u81ea\u5df1\u6559\u5e2b\u3042\u308a\u57fa\u76e4\u30e2\u30c7\u30eb\u306b\u95a2\u3059\u308b\u7cfb\u7d71\u7684\u30ec\u30d3\u30e5\u30fc\uff1a\u8133\u6ce2\u3092\u7528\u3044\u308b<\/h2>\n<ul>\n<li><strong>\u539f\u984c:<\/strong> Systematic review of self-supervised foundation models for brain network representation using electroencephalography<\/li>\n<\/ul>\n<h3 class=\"wp-block-heading\">\u5c02\u9580\u30a2\u30ca\u30ea\u30b9\u30c8\u306e\u5206\u6790<\/h3>\n<div class=\"ai-summary-content\">\n<p>\u672c\u7814\u7a76\u306f\u3001<strong>\u81ea\u5df1\u6559\u5e2b\u3042\u308a\u5b66\u7fd2\uff08SSL\uff09<\/strong>\u3092\u7528\u3044\u305f<strong>\u8133\u6ce2\uff08EEG\uff09\u57fa\u76e4\u30e2\u30c7\u30eb<\/strong>\u306b\u95a2\u3059\u308b\u7cfb\u7d71\u7684\u30ec\u30d3\u30e5\u30fc\u3092\u5b9f\u65bd\u3057\u307e\u3057\u305f\u3002\u3053\u308c\u3089\u306e\u30e2\u30c7\u30eb\u306f\u3001\u5927\u91cf\u306e\u30e9\u30d9\u30eb\u306a\u3057\u30c7\u30fc\u30bf\u3067\u4e8b\u524d\u5b66\u7fd2\u3055\u308c\u3001\u69d8\u3005\u306a\u4e0b\u6d41\u30bf\u30b9\u30af\u306b\u9069\u5fdc\u53ef\u80fd\u3067\u3059\u3002<\/p>\n<p>\u30ec\u30d3\u30e5\u30fc\u3067\u306f\u3001<strong>Transformer\u30a2\u30fc\u30ad\u30c6\u30af\u30c1\u30e3<\/strong>\u304c\u4e3b\u6d41\u3067\u3042\u308a\u3001MAMBA\u3084S4\u306e\u3088\u3046\u306a\u72b6\u614b\u7a7a\u9593\u30e2\u30c7\u30eb\u3082\u767b\u5834\u3057\u3066\u3044\u308b\u3053\u3068\u304c\u7279\u5b9a\u3055\u308c\u307e\u3057\u305f\u3002SSL\u76ee\u7684\u95a2\u6570\u3067\u306f\u3001\u30de\u30b9\u30af\u30c9\u30aa\u30fc\u30c8\u30a8\u30f3\u30b3\u30fc\u30c7\u30a3\u30f3\u30b0\u304c\u6700\u3082\u4e00\u822c\u7684\u3067\u3042\u308a\u3001\u5bfe\u7167\u5b66\u7fd2\u3082\u5229\u7528\u3055\u308c\u3066\u3044\u307e\u3059\u3002<\/p>\n<p>\u3057\u304b\u3057\u3001\u4e8b\u524d\u5b66\u7fd2\u30c7\u30fc\u30bf\u30bb\u30c3\u30c8\u306e\u591a\u69d8\u6027\u304c\u9650\u3089\u308c\u3066\u304a\u308a\u3001\u6a19\u6e96\u5316\u3055\u308c\u305f\u30d9\u30f3\u30c1\u30de\u30fc\u30af\u304c\u5b58\u5728\u3057\u306a\u3044\u3053\u3068\u304c\u8ab2\u984c\u3068\u3057\u3066\u6319\u3052\u3089\u308c\u3066\u3044\u307e\u3059\u3002\u3088\u308a\u5927\u304d\u304f\u591a\u69d8\u306a\u4e8b\u524d\u5b66\u7fd2\u30c7\u30fc\u30bf\u30bb\u30c3\u30c8\u3001\u6a19\u6e96\u5316\u3055\u308c\u305f\u8a55\u4fa1\u30d7\u30ed\u30c8\u30b3\u30eb\u3001\u30de\u30eb\u30c1\u30bf\u30b9\u30af\u691c\u8a3c\u304c\u4eca\u5f8c\u306e\u9032\u6b69\u306b\u4e0d\u53ef\u6b20\u3067\u3059\u3002<\/p>\n<p>\ud83d\udc49 <strong><a href=\"https:\/\/arxiv.org\/abs\/2602.03269\" target=\"_blank\" rel=\"noopener\">arXiv \u3067\u8a18\u4e8b\u5168\u6587\u3092\u8aad\u3080<\/a><\/strong><\/p>\n<\/div>\n<ul>\n<li><strong>\u8981\u70b9:<\/strong> Advancements in EEG foundation models using self-supervised learning show promise but require greater dataset diversity and standardized evaluation for robust, general-purpose applications.<\/li>\n<li><strong>\u8457\u8005:<\/strong> Hannah Portmann, Yosuke Morishima<\/li>\n<\/ul>\n<blockquote class=\"wp-block-quote\"><p><span>English Summary:<\/span><\/p>\n<p>This systematic review examines <strong>self-supervised learning (SSL)<\/strong>-based <strong>electroencephalography (EEG) foundation models<\/strong>, which are pre-trained on large unlabeled datasets and adaptable to various downstream tasks.<\/p>\n<p>The review identified <strong>Transformer architectures<\/strong> as predominant, with emerging alternatives like state-space models such as MAMBA and S4. Masked auto-encoding was the most common SSL objective, with contrastive learning also employed.<\/p>\n<p>Key limitations include the limited diversity of pre-training datasets and the absence of standardized benchmarks. Future progress hinges on larger, more diverse datasets, standardized evaluation protocols, and multi-task validation.<\/p>\n<\/blockquote>\n<\/div>\n<div class=\"wp-block-group\" style=\"margin-top:40px;margin-bottom:40px\">\n<h2 class=\"wp-block-heading\">\u8a00\u8a9e\u30e2\u30c7\u30eb\u306b\u300c\u77e5\u3063\u3066\u3044\u308b\u3053\u3068\u300d\u3092\u8a8d\u8b58\u3055\u305b\u308b\u30d5\u30a1\u30a4\u30f3\u30c1\u30e5\u30fc\u30cb\u30f3\u30b0<\/h2>\n<ul>\n<li><strong>\u539f\u984c:<\/strong> Fine-Tuning Language Models to Know What They Know<\/li>\n<\/ul>\n<h3 class=\"wp-block-heading\">\u5c02\u9580\u30a2\u30ca\u30ea\u30b9\u30c8\u306e\u5206\u6790<\/h3>\n<div class=\"ai-summary-content\">\n<p>\u672c\u7814\u7a76\u3067\u306f\u3001<strong>\u5927\u898f\u6a21\u8a00\u8a9e\u30e2\u30c7\u30eb\uff08LLM\uff09<\/strong>\u306e\u30e1\u30bf\u8a8d\u77e5\u80fd\u529b\u3001\u3059\u306a\u308f\u3061\u81ea\u8eab\u306e\u77e5\u8b58\u72b6\u614b\u3092\u8a8d\u8b58\u3059\u308b\u80fd\u529b\u3092\u5411\u4e0a\u3055\u305b\u308b\u305f\u3081\u306e\u30d5\u30ec\u30fc\u30e0\u30ef\u30fc\u30af\u3092\u63d0\u6848\u3057\u3066\u3044\u307e\u3059\u3002\u63d0\u6848\u624b\u6cd5\u306f\u3001<strong>Evolution Strategy for Metacognitive Alignment (ESMA)<\/strong> \u3068\u547c\u3070\u308c\u3001\u30e2\u30c7\u30eb\u306e\u5185\u90e8\u77e5\u8b58\u3068\u660e\u793a\u7684\u306a\u884c\u52d5\u3092\u9023\u643a\u3055\u305b\u307e\u3059\u3002<\/p>\n<p>ESMA\u306f\u3001\u591a\u69d8\u306a\u672a\u5b66\u7fd2\u8a2d\u5b9a\u306b\u304a\u3044\u3066\u5805\u7262\u306a\u6c4e\u5316\u6027\u80fd\u3092\u793a\u3057\u3001\u30e2\u30c7\u30eb\u304c\u81ea\u8eab\u306e\u77e5\u8b58\u3092\u53c2\u7167\u3059\u308b\u80fd\u529b\u3092\u5411\u4e0a\u3055\u305b\u308b\u3053\u3068\u304c\u78ba\u8a8d\u3055\u308c\u307e\u3057\u305f\u3002\u30d1\u30e9\u30e1\u30fc\u30bf\u5206\u6790\u306b\u3088\u308a\u3001\u3053\u308c\u3089\u306e\u6539\u5584\u306f\u5c11\u6570\u306e\u91cd\u8981\u306a\u30d1\u30e9\u30e1\u30fc\u30bf\u5909\u66f4\u306b\u8d77\u56e0\u3059\u308b\u3053\u3068\u304c\u793a\u5506\u3055\u308c\u3066\u3044\u307e\u3059\u3002<\/p>\n<p>\u3053\u306e\u7814\u7a76\u306f\u3001LLM\u304c\u5358\u306b\u5fdc\u7b54\u3092\u751f\u6210\u3059\u308b\u3060\u3051\u3067\u306a\u304f\u3001\u81ea\u8eab\u306e\u77e5\u8b58\u306e\u78ba\u5b9f\u6027\u3092\u8a55\u4fa1\u3057\u3001\u305d\u308c\u3092\u9069\u5207\u306b\u5229\u7528\u3059\u308b\u80fd\u529b\u3092\u7372\u5f97\u3059\u308b\u305f\u3081\u306e\u91cd\u8981\u306a\u4e00\u6b69\u3068\u306a\u308a\u307e\u3059\u3002<\/p>\n<p>\ud83d\udc49 <strong><a href=\"https:\/\/arxiv.org\/abs\/2602.02605\" target=\"_blank\" rel=\"noopener\">arXiv \u3067\u8a18\u4e8b\u5168\u6587\u3092\u8aad\u3080<\/a><\/strong><\/p>\n<\/div>\n<ul>\n<li><strong>\u8981\u70b9:<\/strong> ESMA effectively enhances LLMs' metacognitive abilities, allowing them to better 'know what they know' and reference their internal knowledge more reliably.<\/li>\n<li><strong>\u8457\u8005:<\/strong> Sangjun Park, Elliot Meyerson, Xin Qiu, Risto Miikkulainen<\/li>\n<\/ul>\n<blockquote class=\"wp-block-quote\"><p><span>English Summary:<\/span><\/p>\n<p>This study proposes a framework to enhance the metacognitive ability of <strong>Large Language Models (LLMs)<\/strong>, specifically their awareness of their own knowledge state. The proposed method, <strong>Evolution Strategy for Metacognitive Alignment (ESMA)<\/strong>, aligns a model's internal knowledge with its explicit behaviors.<\/p>\n<p>ESMA demonstrates robust generalization across diverse untrained settings, indicating an improvement in the model's ability to reference its own knowledge. Parameter analysis suggests these improvements stem from a sparse set of significant modifications.<\/p>\n<p>This work represents a significant step towards enabling LLMs not only to generate responses but also to assess and appropriately utilize their knowledge confidence.<\/p>\n<\/blockquote>\n<\/div>\n<div class=\"wp-block-group\" style=\"margin-top:40px;margin-bottom:40px\">\n<h2 class=\"wp-block-heading\">\u6a5f\u80fd\u7684\u78c1\u6c17\u5171\u9cf4\u753b\u50cf\u6cd5\u3092\u7528\u3044\u305f\u8a8d\u77e5\u8ab2\u984c\u4e2d\u306e\u60c5\u5831\u51e6\u7406\u6307\u6a19\u306e\u63a8\u5b9a<\/h2>\n<ul>\n<li><strong>\u539f\u984c:<\/strong> Estimating measures of information processing during cognitive tasks using functional magnetic resonance imaging<\/li>\n<\/ul>\n<h3 class=\"wp-block-heading\">\u5c02\u9580\u30a2\u30ca\u30ea\u30b9\u30c8\u306e\u5206\u6790<\/h3>\n<div class=\"ai-summary-content\">\n<p>\u672c\u7814\u7a76\u3067\u306f\u3001<strong>\u6a5f\u80fd\u7684\u78c1\u6c17\u5171\u9cf4\u753b\u50cf\u6cd5\uff08fMRI\uff09<\/strong>\u30c7\u30fc\u30bf\u3092\u7528\u3044\u3066\u3001\u8a8d\u77e5\u8ab2\u984c\u4e2d\u306e\u60c5\u5831\u51e6\u7406\u306e\u6307\u6a19\u3092\u63a8\u5b9a\u3059\u308b\u305f\u3081\u306e\u65b0\u3057\u3044\u30d5\u30ec\u30fc\u30e0\u30ef\u30fc\u30af\u3092\u63d0\u6848\u3057\u3066\u3044\u307e\u3059\u3002\u3053\u306e\u30d5\u30ec\u30fc\u30e0\u30ef\u30fc\u30af\u306f\u3001<strong>\u30a2\u30af\u30c6\u30a3\u30d6\u60c5\u5831\u30b9\u30c8\u30ec\u30fc\u30b8\uff08AIS\uff09<\/strong>\u3001<strong>\u4f1d\u9054\u30a8\u30f3\u30c8\u30ed\u30d4\u30fc\uff08TE\uff09<\/strong>\u3001\u304a\u3088\u3073<strong>\u6b63\u5473\u30b7\u30ca\u30b8\u30fc<\/strong>\u3092\u5b9a\u91cf\u5316\u3057\u307e\u3059\u3002<\/p>\n<p>\u7279\u306b\u3001\u9650\u3089\u308c\u305f\u30b5\u30f3\u30d7\u30eb\u30b5\u30a4\u30ba\u3001\u975e\u5b9a\u5e38\u6027\u3001\u304a\u3088\u3073\u30bf\u30b9\u30af\u56fa\u6709\u306e\u6587\u8108\u3068\u3044\u3063\u305ffMRI\u5206\u6790\u306e\u8ab2\u984c\u306b\u5bfe\u51e6\u3059\u308b\u305f\u3081\u3001\u6700\u8fd1\u958b\u767a\u3055\u308c\u305f\u30af\u30ed\u30b9\u76f8\u4e92\u60c5\u5831\u91cf\u306b\u57fa\u3065\u304f\u30a2\u30d7\u30ed\u30fc\u30c1\u304c\u6d3b\u7528\u3055\u308c\u3066\u3044\u307e\u3059\u3002\u3053\u306e\u624b\u6cd5\u306f\u3001\u5b89\u9759\u6642\u304a\u3088\u3073\u30bf\u30b9\u30af\u4e2d\u306e\u30c7\u30fc\u30bf\u3092\u7d44\u307f\u5408\u308f\u305b\u3066\u60c5\u5831\u7406\u8ad6\u7684\u5c3a\u5ea6\u3092\u8a08\u7b97\u3057\u307e\u3059\u3002<\/p>\n<p>\u63d0\u6848\u3055\u308c\u305f\u30d5\u30ec\u30fc\u30e0\u30ef\u30fc\u30af\u306f\u3001\u30d2\u30c8\u30b3\u30cd\u30af\u30c8\u30fc\u30e0\u30d7\u30ed\u30b8\u30a7\u30af\u30c8\u306eN\u30d0\u30c3\u30af\u8ab2\u984c\u30c7\u30fc\u30bf\u306b\u9069\u7528\u3055\u308c\u3001\u30ef\u30fc\u30ad\u30f3\u30b0\u30e1\u30e2\u30ea\u8ca0\u8377\u306e\u5897\u52a0\u306b\u4f34\u3046\u524d\u982d\u982d\u9802\u8449\u9818\u57df\u3067\u306eAIS\u306e\u5897\u52a0\u3001\u5236\u5fa1\u7d4c\u8def\u9593\u306e\u60c5\u5831\u30d5\u30ed\u30fc\u306e\u5897\u5f37\u3092\u793a\u3059TE\u3001\u304a\u3088\u3073\u5168\u4f53\u7684\u306a\u5197\u9577\u6027\u3078\u306e\u30b7\u30d5\u30c8\u3092\u793a\u3059\u6b63\u5473\u30b7\u30ca\u30b8\u30fc\u304c\u89b3\u5bdf\u3055\u308c\u307e\u3057\u305f\u3002<\/p>\n<p>\ud83d\udc49 <strong><a href=\"https:\/\/arxiv.org\/abs\/2602.03240\" target=\"_blank\" rel=\"noopener\">arXiv \u3067\u8a18\u4e8b\u5168\u6587\u3092\u8aad\u3080<\/a><\/strong><\/p>\n<\/div>\n<ul>\n<li><strong>\u8981\u70b9:<\/strong> A novel fMRI analysis framework enables the quantification of information processing measures like AIS and TE, offering new insights into cognitive functions.<\/li>\n<li><strong>\u8457\u8005:<\/strong> Chetan Gohil, Oliver M. Cliff, James M. Shine, Ben D. Fulcher, Joseph T. Lizier<\/li>\n<\/ul>\n<blockquote class=\"wp-block-quote\"><p><span>English Summary:<\/span><\/p>\n<p>This study introduces a novel framework for estimating measures of information processing during cognitive tasks using <strong>functional magnetic resonance imaging (fMRI)<\/strong> data. The framework quantifies <strong>active information storage (AIS)<\/strong>, <strong>transfer entropy (TE)<\/strong>, and <strong>net synergy<\/strong>.<\/p>\n<p>Crucially, it leverages a recently developed cross-mutual information approach to address challenges in fMRI analysis, such as limited sample size, non-stationarity, and task-specific context, by combining resting-state and task data.<\/p>\n<p>Applied to the N-back task from the Human Connectome Project, the framework revealed increased AIS in fronto-parietal regions with working memory load, enhanced directed information flow across control pathways (TE), and a global shift towards redundancy (net synergy).<\/p>\n<\/blockquote>\n<\/div>\n","protected":false},"excerpt":{"rendered":"<p>\u81ea\u5df1\u6559\u5e2b\u3042\u308a\u5b66\u7fd2\u3092\u7528\u3044\u305f\u8133\u30cd\u30c3\u30c8\u30ef\u30fc\u30af\u8868\u73fe\u3001LLM\u306e\u77e5\u8b58\u3092\u5411\u4e0a\u3055\u305b\u308b\u30d5\u30a1\u30a4\u30f3\u30c1\u30e5\u30fc\u30cb\u30f3\u30b0\u624b\u6cd5\u3001fMRI\u3092\u7528\u3044\u305f\u60c5\u5831\u51e6\u7406\u306e\u63a8\u5b9a\u306b\u95a2\u3059\u308b\u6700\u65b0AI\u7814\u7a76\u3092\u89e3\u8aac\u3002<\/p>\n","protected":false},"author":1,"featured_media":849,"comment_status":"","ping_status":"open","sticky":false,"template":"","format":"standard","meta":{"vkexunit_cta_each_option":"","footnotes":""},"categories":[3],"tags":[8,140,16,141,15,139,138],"class_list":{"0":"post-917","1":"post","2":"type-post","3":"status-publish","4":"format-standard","5":"has-post-thumbnail","6":"hentry","7":"category-columns","8":"tag-ai","9":"tag-fmri","10":"tag-llm","11":"tag-141","13":"tag-139","14":"tag-138"},"_links":{"self":[{"href":"https:\/\/itexplore.org\/jp\/wp-json\/wp\/v2\/posts\/917","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/itexplore.org\/jp\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/itexplore.org\/jp\/wp-json\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/itexplore.org\/jp\/wp-json\/wp\/v2\/users\/1"}],"replies":[{"embeddable":true,"href":"https:\/\/itexplore.org\/jp\/wp-json\/wp\/v2\/comments?post=917"}],"version-history":[{"count":0,"href":"https:\/\/itexplore.org\/jp\/wp-json\/wp\/v2\/posts\/917\/revisions"}],"wp:featuredmedia":[{"embeddable":true,"href":"https:\/\/itexplore.org\/jp\/wp-json\/wp\/v2\/media\/849"}],"wp:attachment":[{"href":"https:\/\/itexplore.org\/jp\/wp-json\/wp\/v2\/media?parent=917"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/itexplore.org\/jp\/wp-json\/wp\/v2\/categories?post=917"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/itexplore.org\/jp\/wp-json\/wp\/v2\/tags?post=917"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}