{"id":1124,"date":"2025-02-16T11:52:04","date_gmt":"2025-02-16T11:52:04","guid":{"rendered":"http:\/\/logicalday.site\/?p=1124"},"modified":"2025-02-16T16:50:53","modified_gmt":"2025-02-16T16:50:53","slug":"paper-review-modernbert","status":"publish","type":"post","link":"https:\/\/logicalday.site\/paper-review-modernbert\/","title":{"rendered":"[Paper Review] ModernBERT"},"content":{"rendered":"\n<h1 class=\"wp-block-heading\">&lt;Smarter, Better, Faster, Longer: A Modern Bidirectional Encoder for Fast, Memory Efficient, and Long Context Finetuning and Inference&gt;<\/h1>\n\n\n\n<p><strong>Author<\/strong>: Benjamin Warner et al.<br><strong>Institution<\/strong>: Answer.AI, LightOn, Johns Hopkins University, NVIDIA, HuggingFace<\/p>\n\n\n\n<h2 class=\"wp-block-heading\">Summary<\/h2>\n\n\n\n<ul class=\"wp-block-list\">\n<li>Actually, as other BERT-based models, <strong>ModernBERT<\/strong> model is merely a <strong>newly configurated<\/strong> BERT model with some modifications in model architecture.<\/li>\n\n\n\n<li>Key points are: RoPE, unpadding, OLMo tokenizer, sequence packing, context length extension, etc.<\/li>\n<\/ul>\n\n\n\n<hr class=\"wp-block-separator has-alpha-channel-opacity\"\/>\n\n\n\n<h2 class=\"wp-block-heading has-background\" style=\"border-style:none;border-width:0px;background:linear-gradient(135deg,rgb(255,245,203) 0%,rgb(182,227,212) 21%,rgb(51,167,181) 100%)\">Introduction<\/h2>\n\n\n\n<ul class=\"wp-block-list\">\n<li>Even after the advent of many LLM models, <strong>encoder-only models<\/strong> like BERT are still widely being used in non-generative downstream applications.\n<ul class=\"wp-block-list\">\n<li>Especially in Information Retrieval (IR). Encoder-based semantic search is a core of <strong>RAG (Retrieval-Augmented Generatrion) <\/strong>pipelines.<\/li>\n<\/ul>\n<\/li>\n<\/ul>\n\n\n\n<ul class=\"wp-block-list\">\n<li>However, many pipelines <strong>still rely on older models<\/strong> like original BERT without any adjustment, despite of recent development.<br>So, there are many drawbacks: 512 tokens limit, bad model design and vocab size, bad performance and computational efficiency.<\/li>\n<\/ul>\n\n\n\n<ul class=\"wp-block-list\">\n<li>Of course, many models were developed to overcome this, like these:<\/li>\n<\/ul>\n\n\n\n<div class=\"wp-block-buttons is-content-justification-center is-layout-flex wp-container-core-buttons-is-layout-1 wp-block-buttons-is-layout-flex\">\n<div class=\"wp-block-button\"><a class=\"wp-block-button__link has-text-align-right wp-element-button\">MosaicBERT<\/a><\/div>\n\n\n\n<div class=\"wp-block-button\"><a class=\"wp-block-button__link has-text-align-center wp-element-button\">CrammingBERT<\/a><\/div>\n\n\n\n<div class=\"wp-block-button\"><a class=\"wp-block-button__link has-text-align-center wp-element-button\">AcademicBERT<\/a><\/div>\n\n\n\n<div class=\"wp-block-button\"><a class=\"wp-block-button__link has-text-align-center wp-element-button\">NomicBERT<\/a><\/div>\n\n\n\n<div class=\"wp-block-button\"><a class=\"wp-block-button__link wp-element-button\">GTE-en-MLM<\/a><\/div>\n<\/div>\n\n\n\n<p><\/p>\n\n\n\n<div class=\"wp-block-group is-layout-constrained wp-block-group-is-layout-constrained\">\n<details class=\"wp-block-details is-layout-flow wp-block-details-is-layout-flow\"><summary>This research presented <strong>ModernBERT <\/strong>with the following <strong>contributions<\/strong>:<\/summary>\n<ul class=\"wp-block-list\">\n<li><strong>Improved architecture<\/strong> for increasing downstream performance and efficiency, especially over longer sequence lengths.<\/li>\n\n\n\n<li>Training on modern and <strong>larger scale of data<\/strong> (2 trillion tokens)<\/li>\n\n\n\n<li>Two models: <strong>ModernBERT-base<\/strong> and <strong>ModernBERT-large<\/strong> with SOTA performance.<\/li>\n\n\n\n<li>Result: processing <strong>8192 tokens, two times faster<\/strong> than previous models.<\/li>\n<\/ul>\n<\/details>\n<\/div>\n\n\n\n<hr class=\"wp-block-separator has-alpha-channel-opacity\"\/>\n\n\n\n<h2 class=\"wp-block-heading has-background\" style=\"background:linear-gradient(135deg,rgb(255,245,203) 0%,rgb(182,227,212) 17%,rgb(51,167,181) 100%)\">Methods<\/h2>\n\n\n\n<h3 class=\"wp-block-heading\">Architectural Improvements<\/h3>\n\n\n\n<div class=\"wp-block-group is-layout-constrained wp-block-group-is-layout-constrained\">\n<details class=\"wp-block-details is-layout-flow wp-block-details-is-layout-flow\"><summary>Modern Transformer<\/summary>\n<ul class=\"wp-block-list\">\n<li><strong>Disabled bias terms <\/strong>in all linear layers except for the final decoder linear layer. And disabled all bias terms in Layer Norms too.<\/li>\n<\/ul>\n\n\n\n<ul class=\"wp-block-list\">\n<li>Used<strong> rotary positional embeddings (RoPE)<\/strong>.<\/li>\n<\/ul>\n\n\n\n<ul class=\"wp-block-list\">\n<li><strong>Pre-normalization<\/strong>: Added a LayerNorm after the embedding layer and removed the first LayerNorm in the first attention layer.<\/li>\n<\/ul>\n\n\n\n<ul class=\"wp-block-list\">\n<li><strong>GeGLU <\/strong>activation funcdtion.<\/li>\n<\/ul>\n<\/details>\n<\/div>\n\n\n\n<div class=\"wp-block-group is-layout-constrained wp-block-group-is-layout-constrained\">\n<details class=\"wp-block-details is-layout-flow wp-block-details-is-layout-flow\"><summary>Efficiency Improvements<\/summary>\n<ul class=\"wp-block-list\">\n<li>Attention layers <strong>alternate <\/strong>between <strong>global attention and local attention<\/strong>.<br>(Every 3rd layer is global attention, others are local attentions.)\n<ul class=\"wp-block-list\">\n<li>Global Attention: Every token within a sequence attends to every other token.<\/li>\n\n\n\n<li>Local Attention: Tokens only attend to each other within a small sliding window.<\/li>\n<\/ul>\n<\/li>\n\n\n\n<li><strong>Unpadding<\/strong>: <strong>Removed p<\/strong>adding tokens to avoid wasting computation, concatenated all sequences from a minibatch into a single sequence.<br>Used <strong>Flash Attention&#8217;s variable length attention<\/strong> and RoPE implementations, allowing jagged attention masks. ModernBERT unpads inputs before the token embedding layer and optionally <strong>repads<\/strong> model outputs, leading to performance improvements.<\/li>\n\n\n\n<li><strong>Flash Attention<\/strong>: Provides memory and compute efficient attention kernels.<\/li>\n<\/ul>\n<\/details>\n<\/div>\n\n\n\n<div class=\"wp-block-group is-layout-constrained wp-block-group-is-layout-constrained\">\n<details class=\"wp-block-details is-layout-flow wp-block-details-is-layout-flow\"><summary>Model Design<\/summary>\n<ul class=\"wp-block-list\">\n<li><em><strong>&#8220;Deep &amp; Narrow&#8221;<\/strong><\/em> models are empirically better than <em>&#8220;Shallow &amp; Wide&#8221;<\/em> models, so ModernBERT was designed to be as deep and narrow as possible.<\/li>\n\n\n\n<li>ModernBERT-<strong>base<\/strong>: <strong>22 <\/strong>layers, <strong>149M <\/strong>parameters, <strong>768 <\/strong>hidden size with a GLU expansion of <strong>2,304<\/strong>.<\/li>\n\n\n\n<li>ModernBERT-<strong>large<\/strong>: <strong>28 <\/strong>layers, <strong>395M <\/strong>parameters. <strong>1,024 <\/strong>hidden size, GLU expansion of <strong>5,248<\/strong>.<\/li>\n<\/ul>\n<\/details>\n<\/div>\n\n\n\n<h3 class=\"wp-block-heading\">Training<\/h3>\n\n\n\n<details class=\"wp-block-details has-background is-layout-flow wp-block-details-is-layout-flow\" style=\"background-color:#00000000\"><summary>Tokenizer<\/summary>\n<ul class=\"wp-block-list\">\n<li>Used a <strong>modified version of the OLMo tokenizer<\/strong>, rather than the original BERT tokenizer.<\/li>\n\n\n\n<li>Used the same <strong>special tokens<\/strong> ([CLS] and [SEP]) as the original BERT model.<\/li>\n\n\n\n<li>Vocab size: 50,368<\/li>\n<\/ul>\n\n\n\n<details class=\"wp-block-details is-layout-flow wp-block-details-is-layout-flow\"><summary>Sequence Packing<\/summary>\n<ul class=\"wp-block-list\">\n<li>Because of unpadding, minibatch-size can vary. To avoid high variance, <strong>sequence packing<\/strong> with a greedy algorithm was used. Namely, the algorithm tries to find optimal combination of texts to ensure <strong>uniform batch size<\/strong>.<\/li>\n<\/ul>\n\n\n\n<details class=\"wp-block-details is-layout-flow wp-block-details-is-layout-flow\"><summary>MLM<\/summary>\n<ul class=\"wp-block-list\">\n<li>Followed <strong>MLM (Masked Language Modeling)<\/strong> setup used by <strong>MosaicBERT<\/strong>.<\/li>\n\n\n\n<li>Removed NSP (Next-Sentence Prediction) objective, and used a <strong>masking rate of 30%.<\/strong><\/li>\n<\/ul>\n<\/details>\n<\/details>\n<\/details>\n\n\n\n<details class=\"wp-block-details is-layout-flow wp-block-details-is-layout-flow\"><summary>Batch Size Schedule<\/summary>\n<ul class=\"wp-block-list\">\n<li>Starts with smaller accumulated batches, increasing over time to the full batch size (to accelarate training progress)<\/li>\n<\/ul>\n<\/details>\n\n\n\n<details class=\"wp-block-details is-layout-flow wp-block-details-is-layout-flow\"><summary>Context Length Extension<\/summary>\n<ul class=\"wp-block-list\">\n<li>Thanks to <strong>RoPE<\/strong>, which is <strong>independent <\/strong>of sequence length, it&#8217;s possible to <strong>vary sequence lengths.<\/strong><\/li>\n\n\n\n<li>ModernBERT extended the context length from 1024 tokens to 8192 tokens (after training on 1.7T tokens, and trained on additional 300B tokens).<\/li>\n\n\n\n<li>This enables the model to capture long-range dependencies with training efficiency.<\/li>\n<\/ul>\n<\/details>\n\n\n\n<p>With these settings, <strong>ModernBERT <\/strong>showed the <strong>state-of-the-art performance<\/strong> with the <strong>highest training efficiency.<\/strong><\/p>\n\n\n\n<p><\/p>\n","protected":false},"excerpt":{"rendered":"<p>&lt;Smarter, Better, Faster, Longer: A Modern Bidirectional Encoder for Fast, Memory Efficient, and Long Context Finetuning and Inference&gt; Author: Benjamin Warner et al.Institution: Answer.AI, LightOn, Johns Hopkins University, NVIDIA, HuggingFace Summary Introduction Methods Architectural Improvements Training With these settings, ModernBERT showed the state-of-the-art performance with the highest training efficiency.<\/p>\n","protected":false},"author":1,"featured_media":1128,"comment_status":"open","ping_status":"open","sticky":false,"template":"","format":"standard","meta":{"site-container-style":"default","site-container-layout":"default","site-sidebar-layout":"default","disable-article-header":"default","disable-site-header":"default","disable-site-footer":"default","disable-content-area-spacing":"default","footnotes":""},"categories":[54,53],"tags":[56,55,58],"class_list":["post-1124","post","type-post","status-publish","format-standard","has-post-thumbnail","hentry","category-nlp","category-paper-review","tag-bert","tag-modernbert","tag-modernbert-paper-review"],"yoast_head":"<!-- This site is optimized with the Yoast SEO plugin v24.5 - https:\/\/yoast.com\/wordpress\/plugins\/seo\/ -->\n<title>[Paper Review] ModernBERT<\/title>\n<meta name=\"description\" content=\"Actually, ModernBERT model is merely a newly configurated BERT model with some modifications in model architecture.\" \/>\n<meta name=\"robots\" content=\"index, follow, max-snippet:-1, max-image-preview:large, max-video-preview:-1\" \/>\n<link rel=\"canonical\" href=\"https:\/\/logicalday.site\/paper-review-modernbert\/\" \/>\n<meta property=\"og:locale\" content=\"en_US\" \/>\n<meta property=\"og:type\" content=\"article\" \/>\n<meta property=\"og:title\" content=\"[Paper Review] ModernBERT\" \/>\n<meta property=\"og:description\" content=\"Actually, ModernBERT model is merely a newly configurated BERT model with some modifications in model architecture.\" \/>\n<meta property=\"og:url\" content=\"https:\/\/logicalday.site\/paper-review-modernbert\/\" \/>\n<meta property=\"og:site_name\" content=\"Logical Day\" \/>\n<meta property=\"article:published_time\" content=\"2025-02-16T11:52:04+00:00\" \/>\n<meta property=\"article:modified_time\" content=\"2025-02-16T16:50:53+00:00\" \/>\n<meta property=\"og:image\" content=\"https:\/\/logicalday.site\/wp-content\/uploads\/2025\/02\/modernBERT.png\" \/>\n\t<meta property=\"og:image:width\" content=\"1362\" \/>\n\t<meta property=\"og:image:height\" content=\"762\" \/>\n\t<meta property=\"og:image:type\" content=\"image\/png\" \/>\n<meta name=\"author\" content=\"hankyeol\" \/>\n<meta name=\"twitter:card\" content=\"summary_large_image\" \/>\n<meta name=\"twitter:label1\" content=\"Written by\" \/>\n\t<meta name=\"twitter:data1\" content=\"hankyeol\" \/>\n\t<meta name=\"twitter:label2\" content=\"Est. reading time\" \/>\n\t<meta name=\"twitter:data2\" content=\"3 minutes\" \/>\n<script type=\"application\/ld+json\" class=\"yoast-schema-graph\">{\"@context\":\"https:\/\/schema.org\",\"@graph\":[{\"@type\":\"Article\",\"@id\":\"https:\/\/logicalday.site\/paper-review-modernbert\/#article\",\"isPartOf\":{\"@id\":\"https:\/\/logicalday.site\/paper-review-modernbert\/\"},\"author\":{\"name\":\"hankyeol\",\"@id\":\"https:\/\/logicalday.site\/#\/schema\/person\/75119726e532f4a86b64b03a8acdf17a\"},\"headline\":\"[Paper Review] ModernBERT\",\"datePublished\":\"2025-02-16T11:52:04+00:00\",\"dateModified\":\"2025-02-16T16:50:53+00:00\",\"mainEntityOfPage\":{\"@id\":\"https:\/\/logicalday.site\/paper-review-modernbert\/\"},\"wordCount\":583,\"commentCount\":1,\"publisher\":{\"@id\":\"https:\/\/logicalday.site\/#\/schema\/person\/75119726e532f4a86b64b03a8acdf17a\"},\"image\":{\"@id\":\"https:\/\/logicalday.site\/paper-review-modernbert\/#primaryimage\"},\"thumbnailUrl\":\"https:\/\/logicalday.site\/wp-content\/uploads\/2025\/02\/modernBERT.png\",\"keywords\":[\"BERT\",\"ModernBERT\",\"ModernBERT paper review\"],\"articleSection\":[\"NLP\",\"Paper Review\"],\"inLanguage\":\"en-US\",\"potentialAction\":[{\"@type\":\"CommentAction\",\"name\":\"Comment\",\"target\":[\"https:\/\/logicalday.site\/paper-review-modernbert\/#respond\"]}]},{\"@type\":\"WebPage\",\"@id\":\"https:\/\/logicalday.site\/paper-review-modernbert\/\",\"url\":\"https:\/\/logicalday.site\/paper-review-modernbert\/\",\"name\":\"[Paper Review] ModernBERT\",\"isPartOf\":{\"@id\":\"https:\/\/logicalday.site\/#website\"},\"primaryImageOfPage\":{\"@id\":\"https:\/\/logicalday.site\/paper-review-modernbert\/#primaryimage\"},\"image\":{\"@id\":\"https:\/\/logicalday.site\/paper-review-modernbert\/#primaryimage\"},\"thumbnailUrl\":\"https:\/\/logicalday.site\/wp-content\/uploads\/2025\/02\/modernBERT.png\",\"datePublished\":\"2025-02-16T11:52:04+00:00\",\"dateModified\":\"2025-02-16T16:50:53+00:00\",\"description\":\"Actually, ModernBERT model is merely a newly configurated BERT model with some modifications in model architecture.\",\"breadcrumb\":{\"@id\":\"https:\/\/logicalday.site\/paper-review-modernbert\/#breadcrumb\"},\"inLanguage\":\"en-US\",\"potentialAction\":[{\"@type\":\"ReadAction\",\"target\":[\"https:\/\/logicalday.site\/paper-review-modernbert\/\"]}]},{\"@type\":\"ImageObject\",\"inLanguage\":\"en-US\",\"@id\":\"https:\/\/logicalday.site\/paper-review-modernbert\/#primaryimage\",\"url\":\"https:\/\/logicalday.site\/wp-content\/uploads\/2025\/02\/modernBERT.png\",\"contentUrl\":\"https:\/\/logicalday.site\/wp-content\/uploads\/2025\/02\/modernBERT.png\",\"width\":1362,\"height\":762},{\"@type\":\"BreadcrumbList\",\"@id\":\"https:\/\/logicalday.site\/paper-review-modernbert\/#breadcrumb\",\"itemListElement\":[{\"@type\":\"ListItem\",\"position\":1,\"name\":\"Home\",\"item\":\"https:\/\/logicalday.site\/\"},{\"@type\":\"ListItem\",\"position\":2,\"name\":\"[Paper Review] ModernBERT\"}]},{\"@type\":\"WebSite\",\"@id\":\"https:\/\/logicalday.site\/#website\",\"url\":\"https:\/\/logicalday.site\/\",\"name\":\"Logical Day\",\"description\":\"Study in AI\",\"publisher\":{\"@id\":\"https:\/\/logicalday.site\/#\/schema\/person\/75119726e532f4a86b64b03a8acdf17a\"},\"potentialAction\":[{\"@type\":\"SearchAction\",\"target\":{\"@type\":\"EntryPoint\",\"urlTemplate\":\"https:\/\/logicalday.site\/?s={search_term_string}\"},\"query-input\":{\"@type\":\"PropertyValueSpecification\",\"valueRequired\":true,\"valueName\":\"search_term_string\"}}],\"inLanguage\":\"en-US\"},{\"@type\":[\"Person\",\"Organization\"],\"@id\":\"https:\/\/logicalday.site\/#\/schema\/person\/75119726e532f4a86b64b03a8acdf17a\",\"name\":\"hankyeol\",\"image\":{\"@type\":\"ImageObject\",\"inLanguage\":\"en-US\",\"@id\":\"https:\/\/logicalday.site\/#\/schema\/person\/image\/\",\"url\":\"http:\/\/logicalday.site\/wp-content\/uploads\/2025\/02\/site_icon.png\",\"contentUrl\":\"http:\/\/logicalday.site\/wp-content\/uploads\/2025\/02\/site_icon.png\",\"width\":512,\"height\":512,\"caption\":\"hankyeol\"},\"logo\":{\"@id\":\"https:\/\/logicalday.site\/#\/schema\/person\/image\/\"},\"sameAs\":[\"http:\/\/13.237.90.95\"],\"url\":\"https:\/\/logicalday.site\/author\/hankyeol\/\"}]}<\/script>\n<!-- \/ Yoast SEO plugin. -->","yoast_head_json":{"title":"[Paper Review] ModernBERT","description":"Actually, ModernBERT model is merely a newly configurated BERT model with some modifications in model architecture.","robots":{"index":"index","follow":"follow","max-snippet":"max-snippet:-1","max-image-preview":"max-image-preview:large","max-video-preview":"max-video-preview:-1"},"canonical":"https:\/\/logicalday.site\/paper-review-modernbert\/","og_locale":"en_US","og_type":"article","og_title":"[Paper Review] ModernBERT","og_description":"Actually, ModernBERT model is merely a newly configurated BERT model with some modifications in model architecture.","og_url":"https:\/\/logicalday.site\/paper-review-modernbert\/","og_site_name":"Logical Day","article_published_time":"2025-02-16T11:52:04+00:00","article_modified_time":"2025-02-16T16:50:53+00:00","og_image":[{"width":1362,"height":762,"url":"https:\/\/logicalday.site\/wp-content\/uploads\/2025\/02\/modernBERT.png","type":"image\/png"}],"author":"hankyeol","twitter_card":"summary_large_image","twitter_misc":{"Written by":"hankyeol","Est. reading time":"3 minutes"},"schema":{"@context":"https:\/\/schema.org","@graph":[{"@type":"Article","@id":"https:\/\/logicalday.site\/paper-review-modernbert\/#article","isPartOf":{"@id":"https:\/\/logicalday.site\/paper-review-modernbert\/"},"author":{"name":"hankyeol","@id":"https:\/\/logicalday.site\/#\/schema\/person\/75119726e532f4a86b64b03a8acdf17a"},"headline":"[Paper Review] ModernBERT","datePublished":"2025-02-16T11:52:04+00:00","dateModified":"2025-02-16T16:50:53+00:00","mainEntityOfPage":{"@id":"https:\/\/logicalday.site\/paper-review-modernbert\/"},"wordCount":583,"commentCount":1,"publisher":{"@id":"https:\/\/logicalday.site\/#\/schema\/person\/75119726e532f4a86b64b03a8acdf17a"},"image":{"@id":"https:\/\/logicalday.site\/paper-review-modernbert\/#primaryimage"},"thumbnailUrl":"https:\/\/logicalday.site\/wp-content\/uploads\/2025\/02\/modernBERT.png","keywords":["BERT","ModernBERT","ModernBERT paper review"],"articleSection":["NLP","Paper Review"],"inLanguage":"en-US","potentialAction":[{"@type":"CommentAction","name":"Comment","target":["https:\/\/logicalday.site\/paper-review-modernbert\/#respond"]}]},{"@type":"WebPage","@id":"https:\/\/logicalday.site\/paper-review-modernbert\/","url":"https:\/\/logicalday.site\/paper-review-modernbert\/","name":"[Paper Review] ModernBERT","isPartOf":{"@id":"https:\/\/logicalday.site\/#website"},"primaryImageOfPage":{"@id":"https:\/\/logicalday.site\/paper-review-modernbert\/#primaryimage"},"image":{"@id":"https:\/\/logicalday.site\/paper-review-modernbert\/#primaryimage"},"thumbnailUrl":"https:\/\/logicalday.site\/wp-content\/uploads\/2025\/02\/modernBERT.png","datePublished":"2025-02-16T11:52:04+00:00","dateModified":"2025-02-16T16:50:53+00:00","description":"Actually, ModernBERT model is merely a newly configurated BERT model with some modifications in model architecture.","breadcrumb":{"@id":"https:\/\/logicalday.site\/paper-review-modernbert\/#breadcrumb"},"inLanguage":"en-US","potentialAction":[{"@type":"ReadAction","target":["https:\/\/logicalday.site\/paper-review-modernbert\/"]}]},{"@type":"ImageObject","inLanguage":"en-US","@id":"https:\/\/logicalday.site\/paper-review-modernbert\/#primaryimage","url":"https:\/\/logicalday.site\/wp-content\/uploads\/2025\/02\/modernBERT.png","contentUrl":"https:\/\/logicalday.site\/wp-content\/uploads\/2025\/02\/modernBERT.png","width":1362,"height":762},{"@type":"BreadcrumbList","@id":"https:\/\/logicalday.site\/paper-review-modernbert\/#breadcrumb","itemListElement":[{"@type":"ListItem","position":1,"name":"Home","item":"https:\/\/logicalday.site\/"},{"@type":"ListItem","position":2,"name":"[Paper Review] ModernBERT"}]},{"@type":"WebSite","@id":"https:\/\/logicalday.site\/#website","url":"https:\/\/logicalday.site\/","name":"Logical Day","description":"Study in AI","publisher":{"@id":"https:\/\/logicalday.site\/#\/schema\/person\/75119726e532f4a86b64b03a8acdf17a"},"potentialAction":[{"@type":"SearchAction","target":{"@type":"EntryPoint","urlTemplate":"https:\/\/logicalday.site\/?s={search_term_string}"},"query-input":{"@type":"PropertyValueSpecification","valueRequired":true,"valueName":"search_term_string"}}],"inLanguage":"en-US"},{"@type":["Person","Organization"],"@id":"https:\/\/logicalday.site\/#\/schema\/person\/75119726e532f4a86b64b03a8acdf17a","name":"hankyeol","image":{"@type":"ImageObject","inLanguage":"en-US","@id":"https:\/\/logicalday.site\/#\/schema\/person\/image\/","url":"http:\/\/logicalday.site\/wp-content\/uploads\/2025\/02\/site_icon.png","contentUrl":"http:\/\/logicalday.site\/wp-content\/uploads\/2025\/02\/site_icon.png","width":512,"height":512,"caption":"hankyeol"},"logo":{"@id":"https:\/\/logicalday.site\/#\/schema\/person\/image\/"},"sameAs":["http:\/\/13.237.90.95"],"url":"https:\/\/logicalday.site\/author\/hankyeol\/"}]}},"_links":{"self":[{"href":"https:\/\/logicalday.site\/wp-json\/wp\/v2\/posts\/1124","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/logicalday.site\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/logicalday.site\/wp-json\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/logicalday.site\/wp-json\/wp\/v2\/users\/1"}],"replies":[{"embeddable":true,"href":"https:\/\/logicalday.site\/wp-json\/wp\/v2\/comments?post=1124"}],"version-history":[{"count":16,"href":"https:\/\/logicalday.site\/wp-json\/wp\/v2\/posts\/1124\/revisions"}],"predecessor-version":[{"id":1155,"href":"https:\/\/logicalday.site\/wp-json\/wp\/v2\/posts\/1124\/revisions\/1155"}],"wp:featuredmedia":[{"embeddable":true,"href":"https:\/\/logicalday.site\/wp-json\/wp\/v2\/media\/1128"}],"wp:attachment":[{"href":"https:\/\/logicalday.site\/wp-json\/wp\/v2\/media?parent=1124"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/logicalday.site\/wp-json\/wp\/v2\/categories?post=1124"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/logicalday.site\/wp-json\/wp\/v2\/tags?post=1124"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}