{"id":1659,"date":"2026-02-20T03:18:25","date_gmt":"2026-02-20T03:18:25","guid":{"rendered":"https:\/\/i10xnews.kinsta.cloud\/inside-llms-probing-hidden-traits-biases"},"modified":"2026-02-20T03:18:25","modified_gmt":"2026-02-20T03:18:25","slug":"inside-llms-probing-hidden-traits-biases","status":"publish","type":"post","link":"https:\/\/i10x.ai\/news\/inside-llms-probing-hidden-traits-biases","title":{"rendered":"Inside LLMs: Probing Hidden Traits and Biases | MIT Research"},"content":{"rendered":"\n<h2 class=\"wp-block-heading\">Inside LLMs: Probing Hidden Traits and Biases<\/h2>\n\n\n\n<h3 class=\"wp-block-heading\">\u26a1 Quick Take<\/h3>\n\n\n\n<blockquote class=\"wp-block-quote is-layout-flow wp-block-quote-is-layout-flow\">\n<p class=\"wp-block-paragraph\">Ever wonder what&#8217;s really going on inside those massive language models we rely on every day? A fresh approach from MIT researchers is starting to lift the veil, revealing the subtle biases, shifting moods, and even personality quirks embedded in their internal workings. This isn&#8217;t just another tweak\u2014it&#8217;s a real pivot from poking at AI from the outside to peering right into its core, which feels essential if we&#8217;re ever going to trust these systems with the big stuff.<\/p>\n<\/blockquote>\n\n\n\n<p class=\"wp-block-paragraph\">Summary: Researchers at MIT have come up with a technique to dig into and measure those hidden traits within <strong>LLMs<\/strong>\u2014like biases that aren&#8217;t obvious, emotional tones, or even broader ideas. By looking closely at how the model&#8217;s internal activations fire up, they go past just checking what comes out and start grasping what&#8217;s happening underneath.<\/p>\n\n\n\n<p class=\"wp-block-paragraph\">What happened: We&#8217;ve typically judged these models by their end results, you know\u2014the text they spit out after a prompt. But this method flips that script. It builds something like a high-tech scanner to pick up on internal signals, spotting things like racial bias or a &#8220;political mindset&#8221; simmering away, even when the output doesn&#8217;t scream it.<\/p>\n\n\n\n<p class=\"wp-block-paragraph\">Why it matters now: With LLMs slipping into sensitive areas like healthcare or finance, just skimming the surface for safety won&#8217;t cut it anymore. This points us toward deeper checks\u2014evaluating the model&#8217;s inner workings directly, rather than relying on polished responses. It&#8217;s like laying the groundwork for turning AI interpretability into something solid, almost like a regular engineering task.<\/p>\n\n\n\n<p class=\"wp-block-paragraph\">Who is most affected: Think AI safety folks, the ML engineers tweaking evaluation setups, and those oversight groups in governance. Deep dives like this could soon be mandatory for audits, pushing creators to show their models aren&#8217;t just behaving right, but thinking that way too\u2014from the inside out.<\/p>\n\n\n\n<p class=\"wp-block-paragraph\">The under-reported angle: Sure, uncovering bias is huge, but this feels like the dawn of &#8220;AI psychology&#8221; in a way. That said, getting from a smart paper to tools that work in the real world? That&#8217;s a steep climb. The tough part will be nailing down standards for these probes, stacking them up against things like activation steering, and handing engineers a reliable guide so they don&#8217;t misread the signals.<\/p>\n\n\n\n<h3 class=\"wp-block-heading\">\ud83e\udde0 Deep Dive<\/h3>\n\n\n\n<p class=\"wp-block-paragraph\">Have you ever felt like we&#8217;re just scratching the surface with AI safety? For so long, the big headache has been that &#8220;black box&#8221; nature of these models. We feed in data, get text back, but the zillions of parameters churning away? Total mystery, like trying to guess the thoughts of some vast, otherworldly brain. Most checks right now are all about behavior\u2014throw prompts at it, cross your fingers it doesn&#8217;t go off the rails with bias or toxicity. Now, MIT researchers are rolling out a method that swaps that hit-or-miss style for something sharper, almost like hooking up an fMRI to the model&#8217;s mind.<\/p>\n\n\n\n<p class=\"wp-block-paragraph\">At its heart, it&#8217;s about probing representations\u2014spotting those telltale patterns in how neurons light up to capture big ideas, from biases to personality vibes. Train a basic probe on one spot in the model to ID a concept, say a certain bias or trait, and boom\u2014you can roll it out to scan everywhere else. What they&#8217;ve found is that <strong>LLMs<\/strong> don&#8217;t just process these things once; they weave them through their whole setup, quietly shaping outputs in ways that catch you off guard. From what I&#8217;ve seen in similar work, this gets us closer to unpacking not only what an AI says, but the reasons behind it\u2014why it picks those words over others.<\/p>\n\n\n\n<p class=\"wp-block-paragraph\">But here&#8217;s the thing: this idea doesn&#8217;t float alone; it needs to tie into the wider world of interpretability. Stuff like <strong>concept activation vectors<\/strong> or <strong>activation steering<\/strong> has been poking around here too. The missing piece this paper leaves hanging? Solid metrics. How does it stack up on precision, toughness against tweaks, or how much compute it guzzles? To move beyond intriguing lab results, it&#8217;ll have to outshine or mesh with what&#8217;s already out there\u2014in direct tests, no less.<\/p>\n\n\n\n<p class=\"wp-block-paragraph\">For the folks in the trenches\u2014ML engineers, safety leads juggling deadlines\u2014this is still more tease than toolkit. No open code, no easy benchmarks, nothing like a plug-and-play guide for weaving it into your deployment flows. Without that practical roadmap, these insights might just gather dust on shelves. Bridging from theory to something you version-control and run in production? It&#8217;s a winding road, full of pitfalls around stats that hold up and whether it works across models like Llama, Gemini, or Claude.<\/p>\n\n\n\n<p class=\"wp-block-paragraph\">And then there&#8217;s the ethical side, which doesn&#8217;t get enough airtime yet. Labeling a bunch of math as having &#8220;mood&#8221; or &#8220;personality&#8221;\u2014that&#8217;s a bold step, philosophically speaking. What even counts as a machine feeling &#8220;cheerful&#8221; or edging toward &#8220;cynical&#8221;? Lean too hard into humanizing it, and you risk muddying the waters, maybe even opening doors to bad calls or abuse. We need clear rules here: what we&#8217;re actually measuring, its blind spots, and guidelines for how these model &#8220;profiles&#8221; fit into audits or regs\u2014before it all spirals.<\/p>\n\n\n\n<h3 class=\"wp-block-heading\">\ud83d\udcca Stakeholders &amp; Impact<\/h3>\n\n\n\n<figure class=\"wp-block-table\"><table class=\"has-fixed-layout\"><thead><tr><th>Stakeholder \/ Aspect<\/th><th>Impact<\/th><th>Insight<\/th><\/tr><\/thead><tbody><tr><td>AI \/ LLM Providers<\/td><td>High<\/td><td>It&#8217;s a way forward for stronger safety pitches and alignment, but it cranks up the pressure on being open about internals. Showing a model&#8217;s &#8220;mind&#8221; is unbiased? Way trickier than tweaking outputs on the fly.<\/td><\/tr><tr><td><strong>MLOps<\/strong> &amp; Tooling Vendors<\/td><td>Medium<\/td><td>Opens doors for fresh tools in monitoring\u2014think seamless probing baked into red-teaming or eval suites, creating real market buzz.<\/td><\/tr><tr><td>Developers &amp; ML Engineers<\/td><td>High<\/td><td>Safety duties evolve from quick fixes after the fact to upfront peeks inside; that means picking up interpretability chops and stats know-how.<\/td><\/tr><tr><td>Regulators &amp; Policy<\/td><td>Significant<\/td><td>Paves the way for audits demanding proof of fair thinking, not just fair talking\u2014could anchor those explainable AI rules down the line.<\/td><\/tr><\/tbody><\/table><\/figure>\n\n\n\n<h3 class=\"wp-block-heading\">\u270d\ufe0f About the analysis<\/h3>\n\n\n\n<p class=\"wp-block-paragraph\">This draws from an independent i10x look at the latest papers on LLM interpretability and bias hunting. I pieced the takeaways by weighing academic breakthroughs against what real-world AI teams need\u2014developers, engineers, and CTOs pushing safe systems live, day to day.<\/p>\n\n\n\n<h3 class=\"wp-block-heading\">\ud83d\udd2d i10x Perspective<\/h3>\n\n\n\n<p class=\"wp-block-paragraph\">From where I stand, this work marks a turning point, nudging AI safety out of fuzzy guesswork and into something more like precise science. We&#8217;re crafting tools to glimpse those emergent inner realms in big models, and it makes you think\u2014the next ten years won&#8217;t just be about bulking up scale, but charting these unseen mental maps.<\/p>\n\n\n\n<p class=\"wp-block-paragraph\">The real pull, though, is that core question: can we <strong>keep up with decoding and steering these beasts<\/strong> as fast as we build them? We&#8217;re getting better at reading the AI&#8217;s &#8220;mind,&#8221; sure\u2014but reshaping it? That&#8217;s still a ways off, plenty of reasons to stay vigilant.<\/p>\n","protected":false},"excerpt":{"rendered":"<p>Explore MIT&#8217;s innovative technique to uncover hidden biases, moods, and personality traits in large language models (LLMs). Dive beyond surface outputs to analyze internal activations for enhanced AI safety and interpretability. Discover how this advances trustworthy AI.<\/p>\n","protected":false},"author":2,"featured_media":1655,"comment_status":"open","ping_status":"open","sticky":false,"template":"","format":"standard","meta":{"paper_title":"","paper_url":"","pdf_url":"","doi":"","arxiv_id":"","authors":"","institutions":"","paper_date":"","venue":"","peer_reviewed":false,"code_url":"","data_url":"","funding_coi":"","figure_license":"","paper_license":"","replication":"","study_models":"","study_sample":"","footnotes":"","_yoast_wpseo_primary_category":""},"categories":[9],"tags":[],"companies":[],"products":[],"class_list":["post-1659","post","type-post","status-publish","format-standard","has-post-thumbnail","hentry","category-models"],"yoast_head":"<!-- This site is optimized with the Yoast SEO plugin v28.7 - https:\/\/yoast.com\/product\/yoast-seo-wordpress\/ -->\n<title>Inside LLMs: Probing Hidden Traits and Biases | MIT Research - i10X News<\/title>\n<meta name=\"description\" content=\"MIT researchers use a new technique to probe hidden biases, moods, and personality traits in LLMs by analyzing internal activations for safer AI.\" \/>\n<meta name=\"robots\" content=\"index, follow, max-snippet:-1, max-image-preview:large, max-video-preview:-1\" \/>\n<link rel=\"canonical\" href=\"https:\/\/i10x.ai\/news\/inside-llms-probing-hidden-traits-biases\" \/>\n<meta property=\"og:locale\" content=\"en_US\" \/>\n<meta property=\"og:type\" content=\"article\" \/>\n<meta property=\"og:title\" content=\"Inside LLMs: Probing Hidden Traits and Biases | MIT Research - i10X News\" \/>\n<meta property=\"og:description\" content=\"MIT researchers use a new technique to probe hidden biases, moods, and personality traits in LLMs by analyzing internal activations for safer AI.\" \/>\n<meta property=\"og:url\" content=\"https:\/\/i10x.ai\/news\/inside-llms-probing-hidden-traits-biases\" \/>\n<meta property=\"og:site_name\" content=\"i10X News\" \/>\n<meta property=\"article:published_time\" content=\"2026-02-20T03:18:25+00:00\" \/>\n<meta property=\"og:image\" content=\"https:\/\/i10x.ai\/news\/wp-content\/uploads\/2026\/10\/probing-llms-biases.webp\" \/>\n\t<meta property=\"og:image:width\" content=\"1344\" \/>\n\t<meta property=\"og:image:height\" content=\"768\" \/>\n\t<meta property=\"og:image:type\" content=\"image\/webp\" \/>\n<meta name=\"author\" content=\"Christopher Ort\" \/>\n<meta name=\"twitter:card\" content=\"summary_large_image\" \/>\n<meta name=\"twitter:creator\" content=\"@i10x_ai\" \/>\n<meta name=\"twitter:site\" content=\"@i10x_ai\" \/>\n<meta name=\"twitter:label1\" content=\"Written by\" \/>\n\t<meta name=\"twitter:data1\" content=\"Christopher Ort\" \/>\n\t<meta name=\"twitter:label2\" content=\"Est. reading time\" \/>\n\t<meta name=\"twitter:data2\" content=\"6 minutes\" \/>\n<script type=\"application\/ld+json\" class=\"yoast-schema-graph\">{\"@context\":\"https:\\\/\\\/schema.org\",\"@graph\":[{\"@type\":\"NewsArticle\",\"@id\":\"https:\\\/\\\/i10x.ai\\\/news\\\/inside-llms-probing-hidden-traits-biases#article\",\"isPartOf\":{\"@id\":\"https:\\\/\\\/i10x.ai\\\/news\\\/inside-llms-probing-hidden-traits-biases\"},\"author\":{\"name\":\"Christopher Ort\",\"@id\":\"https:\\\/\\\/i10x.ai\\\/news\\\/#\\\/schema\\\/person\\\/d410bb7ee72febefca2c9c369939219a\"},\"headline\":\"Inside LLMs: Probing Hidden Traits and Biases | MIT Research\",\"datePublished\":\"2026-02-20T03:18:25+00:00\",\"mainEntityOfPage\":{\"@id\":\"https:\\\/\\\/i10x.ai\\\/news\\\/inside-llms-probing-hidden-traits-biases\"},\"wordCount\":1184,\"commentCount\":0,\"publisher\":{\"@id\":\"https:\\\/\\\/i10x.ai\\\/news\\\/#organization\"},\"image\":{\"@id\":\"https:\\\/\\\/i10x.ai\\\/news\\\/inside-llms-probing-hidden-traits-biases#primaryimage\"},\"thumbnailUrl\":\"https:\\\/\\\/i10x.ai\\\/news\\\/wp-content\\\/uploads\\\/2026\\\/10\\\/probing-llms-biases.webp\",\"articleSection\":[\"Models\"],\"inLanguage\":\"en-US\",\"potentialAction\":[{\"@type\":\"CommentAction\",\"name\":\"Comment\",\"target\":[\"https:\\\/\\\/i10x.ai\\\/news\\\/inside-llms-probing-hidden-traits-biases#respond\"]}],\"dateModified\":\"2026-02-20T03:18:25+00:00\"},{\"@type\":\"WebPage\",\"@id\":\"https:\\\/\\\/i10x.ai\\\/news\\\/inside-llms-probing-hidden-traits-biases\",\"url\":\"https:\\\/\\\/i10x.ai\\\/news\\\/inside-llms-probing-hidden-traits-biases\",\"name\":\"Inside LLMs: Probing Hidden Traits and Biases | MIT Research - i10X News\",\"isPartOf\":{\"@id\":\"https:\\\/\\\/i10x.ai\\\/news\\\/#website\"},\"primaryImageOfPage\":{\"@id\":\"https:\\\/\\\/i10x.ai\\\/news\\\/inside-llms-probing-hidden-traits-biases#primaryimage\"},\"image\":{\"@id\":\"https:\\\/\\\/i10x.ai\\\/news\\\/inside-llms-probing-hidden-traits-biases#primaryimage\"},\"thumbnailUrl\":\"https:\\\/\\\/i10x.ai\\\/news\\\/wp-content\\\/uploads\\\/2026\\\/10\\\/probing-llms-biases.webp\",\"datePublished\":\"2026-02-20T03:18:25+00:00\",\"description\":\"MIT researchers use a new technique to probe hidden biases, moods, and personality traits in LLMs by analyzing internal activations for safer AI.\",\"breadcrumb\":{\"@id\":\"https:\\\/\\\/i10x.ai\\\/news\\\/inside-llms-probing-hidden-traits-biases#breadcrumb\"},\"inLanguage\":\"en-US\",\"potentialAction\":[{\"@type\":\"ReadAction\",\"target\":[\"https:\\\/\\\/i10x.ai\\\/news\\\/inside-llms-probing-hidden-traits-biases\"]}]},{\"@type\":\"ImageObject\",\"inLanguage\":\"en-US\",\"@id\":\"https:\\\/\\\/i10x.ai\\\/news\\\/inside-llms-probing-hidden-traits-biases#primaryimage\",\"url\":\"https:\\\/\\\/i10x.ai\\\/news\\\/wp-content\\\/uploads\\\/2026\\\/10\\\/probing-llms-biases.webp\",\"contentUrl\":\"https:\\\/\\\/i10x.ai\\\/news\\\/wp-content\\\/uploads\\\/2026\\\/10\\\/probing-llms-biases.webp\",\"width\":1344,\"height\":768,\"caption\":\"Abstract representation of a complex, glowing blue sphere with interconnected nodes, surrounded by orbiting rings and energy bursts, symbolizing AI internal workings.\"},{\"@type\":\"BreadcrumbList\",\"@id\":\"https:\\\/\\\/i10x.ai\\\/news\\\/inside-llms-probing-hidden-traits-biases#breadcrumb\",\"itemListElement\":[{\"@type\":\"ListItem\",\"position\":1,\"name\":\"Home\",\"item\":\"https:\\\/\\\/i10x.ai\\\/news\"},{\"@type\":\"ListItem\",\"position\":2,\"name\":\"Inside LLMs: Probing Hidden Traits and Biases | MIT Research\"}]},{\"@type\":\"WebSite\",\"@id\":\"https:\\\/\\\/i10x.ai\\\/news\\\/#website\",\"url\":\"https:\\\/\\\/i10x.ai\\\/news\",\"name\":\"i10X News\",\"description\":\"Stay updated with the latest news and announcements from i10X.\",\"publisher\":{\"@id\":\"https:\\\/\\\/i10x.ai\\\/news\\\/#organization\"},\"potentialAction\":[{\"@type\":\"SearchAction\",\"target\":{\"@type\":\"EntryPoint\",\"urlTemplate\":\"https:\\\/\\\/i10x.ai\\\/news?s={search_term_string}\"},\"query-input\":{\"@type\":\"PropertyValueSpecification\",\"valueRequired\":true,\"valueName\":\"search_term_string\"}}],\"inLanguage\":\"en-US\"},{\"@type\":\"Organization\",\"@id\":\"https:\\\/\\\/i10x.ai\\\/news\\\/#organization\",\"name\":\"i10X\",\"url\":\"https:\\\/\\\/i10x.ai\\\/news\\\/\",\"logo\":{\"@type\":\"ImageObject\",\"inLanguage\":\"en-US\",\"@id\":\"https:\\\/\\\/i10x.ai\\\/news\\\/#\\\/schema\\\/logo\\\/image\\\/\",\"url\":\"https:\\\/\\\/i10xnews.kinsta.cloud\\\/wp-content\\\/uploads\\\/2026\\\/10\\\/i10x-logo-180.png\",\"contentUrl\":\"https:\\\/\\\/i10xnews.kinsta.cloud\\\/wp-content\\\/uploads\\\/2026\\\/10\\\/i10x-logo-180.png\",\"width\":180,\"height\":180,\"caption\":\"i10X\"},\"image\":{\"@id\":\"https:\\\/\\\/i10x.ai\\\/news\\\/#\\\/schema\\\/logo\\\/image\\\/\"},\"sameAs\":[\"https:\\\/\\\/x.com\\\/i10x_ai\",\"https:\\\/\\\/www.linkedin.com\\\/company\\\/i10x-ai\",\"https:\\\/\\\/instagram.com\\\/i10x.ai\",\"https:\\\/\\\/www.youtube.com\\\/@i10X_ai_official\"]},{\"@type\":\"Person\",\"@id\":\"https:\\\/\\\/i10x.ai\\\/news\\\/#\\\/schema\\\/person\\\/d410bb7ee72febefca2c9c369939219a\",\"name\":\"Christopher Ort\",\"image\":{\"@type\":\"ImageObject\",\"inLanguage\":\"en-US\",\"@id\":\"https:\\\/\\\/i10x.ai\\\/news\\\/wp-content\\\/uploads\\\/2026\\\/10\\\/christopher-ort-192.webp\",\"url\":\"https:\\\/\\\/i10x.ai\\\/news\\\/wp-content\\\/uploads\\\/2026\\\/10\\\/christopher-ort-192.webp\",\"contentUrl\":\"https:\\\/\\\/i10x.ai\\\/news\\\/wp-content\\\/uploads\\\/2026\\\/10\\\/christopher-ort-192.webp\",\"caption\":\"Christopher Ort\"},\"description\":\"Christopher Ort handles SEO, AI agent workflows and PR & partnerships at i10x.ai, the all-in-one AI workspace built by i10X Global Pte. Ltd. in Singapore. He is a graduate of Friedrich-Alexander-Universit\u00e4t Erlangen-N\u00fcrnberg (FAU). He is the author of \u201cThe Style Penalty\u201d, a study published on SSRN that analyses 1,576 evaluation decisions by four frontier AI models to show how AI-written CVs fare in automated hiring screens; the research has been covered by heise online and t3n.\",\"url\":\"https:\\\/\\\/i10x.ai\\\/news\\\/author\\\/christopher-ort\",\"sameAs\":[\"https:\\\/\\\/i10x.ai\\\/blog\\\/author\\\/christopher-ort\"],\"worksFor\":{\"@id\":\"https:\\\/\\\/i10x.ai\\\/news\\\/#organization\"},\"alumniOf\":{\"@type\":\"EducationalOrganization\",\"name\":\"Friedrich-Alexander-Universit\u00e4t Erlangen-N\u00fcrnberg (FAU)\",\"url\":\"https:\\\/\\\/www.fau.eu\"}}]}<\/script>\n<!-- \/ Yoast SEO plugin. -->","yoast_head_json":{"title":"Inside LLMs: Probing Hidden Traits and Biases | MIT Research - i10X News","description":"MIT researchers use a new technique to probe hidden biases, moods, and personality traits in LLMs by analyzing internal activations for safer AI.","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:\/\/i10x.ai\/news\/inside-llms-probing-hidden-traits-biases","og_locale":"en_US","og_type":"article","og_title":"Inside LLMs: Probing Hidden Traits and Biases | MIT Research - i10X News","og_description":"MIT researchers use a new technique to probe hidden biases, moods, and personality traits in LLMs by analyzing internal activations for safer AI.","og_url":"https:\/\/i10x.ai\/news\/inside-llms-probing-hidden-traits-biases","og_site_name":"i10X News","article_published_time":"2026-02-20T03:18:25+00:00","og_image":[{"width":1344,"height":768,"url":"https:\/\/i10x.ai\/news\/wp-content\/uploads\/2026\/10\/probing-llms-biases.webp","type":"image\/webp"}],"author":"Christopher Ort","twitter_card":"summary_large_image","twitter_creator":"@i10x_ai","twitter_site":"@i10x_ai","twitter_misc":{"Written by":"Christopher Ort","Est. reading time":"6 minutes"},"schema":{"@context":"https:\/\/schema.org","@graph":[{"@type":"NewsArticle","@id":"https:\/\/i10x.ai\/news\/inside-llms-probing-hidden-traits-biases#article","isPartOf":{"@id":"https:\/\/i10x.ai\/news\/inside-llms-probing-hidden-traits-biases"},"author":{"name":"Christopher Ort","@id":"https:\/\/i10x.ai\/news\/#\/schema\/person\/d410bb7ee72febefca2c9c369939219a"},"headline":"Inside LLMs: Probing Hidden Traits and Biases | MIT Research","datePublished":"2026-02-20T03:18:25+00:00","mainEntityOfPage":{"@id":"https:\/\/i10x.ai\/news\/inside-llms-probing-hidden-traits-biases"},"wordCount":1184,"commentCount":0,"publisher":{"@id":"https:\/\/i10x.ai\/news\/#organization"},"image":{"@id":"https:\/\/i10x.ai\/news\/inside-llms-probing-hidden-traits-biases#primaryimage"},"thumbnailUrl":"https:\/\/i10x.ai\/news\/wp-content\/uploads\/2026\/10\/probing-llms-biases.webp","articleSection":["Models"],"inLanguage":"en-US","potentialAction":[{"@type":"CommentAction","name":"Comment","target":["https:\/\/i10x.ai\/news\/inside-llms-probing-hidden-traits-biases#respond"]}],"dateModified":"2026-02-20T03:18:25+00:00"},{"@type":"WebPage","@id":"https:\/\/i10x.ai\/news\/inside-llms-probing-hidden-traits-biases","url":"https:\/\/i10x.ai\/news\/inside-llms-probing-hidden-traits-biases","name":"Inside LLMs: Probing Hidden Traits and Biases | MIT Research - i10X News","isPartOf":{"@id":"https:\/\/i10x.ai\/news\/#website"},"primaryImageOfPage":{"@id":"https:\/\/i10x.ai\/news\/inside-llms-probing-hidden-traits-biases#primaryimage"},"image":{"@id":"https:\/\/i10x.ai\/news\/inside-llms-probing-hidden-traits-biases#primaryimage"},"thumbnailUrl":"https:\/\/i10x.ai\/news\/wp-content\/uploads\/2026\/10\/probing-llms-biases.webp","datePublished":"2026-02-20T03:18:25+00:00","description":"MIT researchers use a new technique to probe hidden biases, moods, and personality traits in LLMs by analyzing internal activations for safer AI.","breadcrumb":{"@id":"https:\/\/i10x.ai\/news\/inside-llms-probing-hidden-traits-biases#breadcrumb"},"inLanguage":"en-US","potentialAction":[{"@type":"ReadAction","target":["https:\/\/i10x.ai\/news\/inside-llms-probing-hidden-traits-biases"]}]},{"@type":"ImageObject","inLanguage":"en-US","@id":"https:\/\/i10x.ai\/news\/inside-llms-probing-hidden-traits-biases#primaryimage","url":"https:\/\/i10x.ai\/news\/wp-content\/uploads\/2026\/10\/probing-llms-biases.webp","contentUrl":"https:\/\/i10x.ai\/news\/wp-content\/uploads\/2026\/10\/probing-llms-biases.webp","width":1344,"height":768,"caption":"Abstract representation of a complex, glowing blue sphere with interconnected nodes, surrounded by orbiting rings and energy bursts, symbolizing AI internal workings."},{"@type":"BreadcrumbList","@id":"https:\/\/i10x.ai\/news\/inside-llms-probing-hidden-traits-biases#breadcrumb","itemListElement":[{"@type":"ListItem","position":1,"name":"Home","item":"https:\/\/i10x.ai\/news"},{"@type":"ListItem","position":2,"name":"Inside LLMs: Probing Hidden Traits and Biases | MIT Research"}]},{"@type":"WebSite","@id":"https:\/\/i10x.ai\/news\/#website","url":"https:\/\/i10x.ai\/news","name":"i10X News","description":"Stay updated with the latest news and announcements from i10X.","publisher":{"@id":"https:\/\/i10x.ai\/news\/#organization"},"potentialAction":[{"@type":"SearchAction","target":{"@type":"EntryPoint","urlTemplate":"https:\/\/i10x.ai\/news?s={search_term_string}"},"query-input":{"@type":"PropertyValueSpecification","valueRequired":true,"valueName":"search_term_string"}}],"inLanguage":"en-US"},{"@type":"Organization","@id":"https:\/\/i10x.ai\/news\/#organization","name":"i10X","url":"https:\/\/i10x.ai\/news\/","logo":{"@type":"ImageObject","inLanguage":"en-US","@id":"https:\/\/i10x.ai\/news\/#\/schema\/logo\/image\/","url":"https:\/\/i10xnews.kinsta.cloud\/wp-content\/uploads\/2026\/10\/i10x-logo-180.png","contentUrl":"https:\/\/i10xnews.kinsta.cloud\/wp-content\/uploads\/2026\/10\/i10x-logo-180.png","width":180,"height":180,"caption":"i10X"},"image":{"@id":"https:\/\/i10x.ai\/news\/#\/schema\/logo\/image\/"},"sameAs":["https:\/\/x.com\/i10x_ai","https:\/\/www.linkedin.com\/company\/i10x-ai","https:\/\/instagram.com\/i10x.ai","https:\/\/www.youtube.com\/@i10X_ai_official"]},{"@type":"Person","@id":"https:\/\/i10x.ai\/news\/#\/schema\/person\/d410bb7ee72febefca2c9c369939219a","name":"Christopher Ort","image":{"@type":"ImageObject","inLanguage":"en-US","@id":"https:\/\/i10x.ai\/news\/wp-content\/uploads\/2026\/10\/christopher-ort-192.webp","url":"https:\/\/i10x.ai\/news\/wp-content\/uploads\/2026\/10\/christopher-ort-192.webp","contentUrl":"https:\/\/i10x.ai\/news\/wp-content\/uploads\/2026\/10\/christopher-ort-192.webp","caption":"Christopher Ort"},"description":"Christopher Ort handles SEO, AI agent workflows and PR & partnerships at i10x.ai, the all-in-one AI workspace built by i10X Global Pte. Ltd. in Singapore. He is a graduate of Friedrich-Alexander-Universit\u00e4t Erlangen-N\u00fcrnberg (FAU). He is the author of \u201cThe Style Penalty\u201d, a study published on SSRN that analyses 1,576 evaluation decisions by four frontier AI models to show how AI-written CVs fare in automated hiring screens; the research has been covered by heise online and t3n.","url":"https:\/\/i10x.ai\/news\/author\/christopher-ort","sameAs":["https:\/\/i10x.ai\/blog\/author\/christopher-ort"],"worksFor":{"@id":"https:\/\/i10x.ai\/news\/#organization"},"alumniOf":{"@type":"EducationalOrganization","name":"Friedrich-Alexander-Universit\u00e4t Erlangen-N\u00fcrnberg (FAU)","url":"https:\/\/www.fau.eu"}}]}},"_links":{"self":[{"href":"https:\/\/i10x.ai\/news\/wp-json\/wp\/v2\/posts\/1659","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/i10x.ai\/news\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/i10x.ai\/news\/wp-json\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/i10x.ai\/news\/wp-json\/wp\/v2\/users\/2"}],"replies":[{"embeddable":true,"href":"https:\/\/i10x.ai\/news\/wp-json\/wp\/v2\/comments?post=1659"}],"version-history":[{"count":1,"href":"https:\/\/i10x.ai\/news\/wp-json\/wp\/v2\/posts\/1659\/revisions"}],"predecessor-version":[{"id":4702,"href":"https:\/\/i10x.ai\/news\/wp-json\/wp\/v2\/posts\/1659\/revisions\/4702"}],"wp:featuredmedia":[{"embeddable":true,"href":"https:\/\/i10x.ai\/news\/wp-json\/wp\/v2\/media\/1655"}],"wp:attachment":[{"href":"https:\/\/i10x.ai\/news\/wp-json\/wp\/v2\/media?parent=1659"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/i10x.ai\/news\/wp-json\/wp\/v2\/categories?post=1659"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/i10x.ai\/news\/wp-json\/wp\/v2\/tags?post=1659"},{"taxonomy":"company","embeddable":true,"href":"https:\/\/i10x.ai\/news\/wp-json\/wp\/v2\/companies?post=1659"},{"taxonomy":"product","embeddable":true,"href":"https:\/\/i10x.ai\/news\/wp-json\/wp\/v2\/products?post=1659"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}