--- title: Exposing biases, moods, personalities, and abstract... description: MIT and UC San Diego researchers have developed a method to detect and manipulate hidden abstract concepts—such as biases, moods, personalities, and... canonical: https://daily.dev/posts/exposing-biases-moods-personalities-and-abstract-concepts-hidden-in-large-language-models-eyfxmvhsd twitter:card: summary_large_image twitter:site: @dailydotdev og:type: website og:site_name: daily.dev og:title: Exposing biases, moods, personalities, and abstract concepts hidden in large language models | daily.dev og:description: MIT and UC San Diego researchers have developed a method to detect and manipulate hidden abstract concepts—such as biases, moods, personalities, and... og:url: https://daily.dev/posts/exposing-biases-moods-personalities-and-abstract-concepts-hidden-in-large-language-models-eyfxmvhsd og:image: https://api.daily.dev/og/posts/EYFxMvhsd.png og:image:alt: Exposing biases, moods, personalities, and abstract concepts hidden in large language models og:image:width: 1200 og:image:height: 630 og:locale: en --- > ## Documentation Index > Fetch the complete documentation index at: https://daily.dev/llms.txt > Use this file to discover all available pages before exploring further. # Exposing biases, moods, personalities, and abstract concepts hidden in large language models **[MIT News](https://daily.dev/sources/mit)** · 7 min read · 0 upvotes · 0 comments ## Summary MIT and UC San Diego researchers have developed a method to detect and manipulate hidden abstract concepts—such as biases, moods, personalities, and fears—within large language models. Using a recursive feature machine (RFM) algorithm, the approach trains on labeled prompts to identify numerical patterns in LLM representations associated with a specific concept, then steers model outputs by amplifying or suppressing those patterns. The team successfully identified and manipulated over 500 concepts across five categories (fears, expert personas, moods, location preferences, and named personas) in several major LLMs. Demonstrations included generating conspiracy-theorist-toned responses and bypassing safety refusals. The researchers see the method as a tool for improving LLM safety and enabling highly specialized models, and have released the underlying code publicly. Findings are published in Science. ## Full article daily.dev links to this article rather than hosting it. Read it at the original source: ## Similar posts on daily.dev - [Why LLMs Aren’t a One-Size-Fits-All Solution for Enterprises](https://daily.dev/posts/why-llms-aren-t-a-one-size-fits-all-solution-for-enterprises-xhbiq4sz7) · Towards Data Science · 0 upvotes · 0 comments - [Researchers discover a shortcoming that makes LLMs less reliable](https://daily.dev/posts/researchers-discover-a-shortcoming-that-makes-llms-less-reliable-wiiiapmcn) · MIT News · 3 upvotes · 0 comments - [Giving LLMs a personality is just good engineering](https://daily.dev/posts/giving-llms-a-personality-is-just-good-engineering-iel6aulal) · sean goedecke · 24 upvotes · 4 comments --- Tags: [#llm](https://daily.dev/tags/llm), [#neural-networks](https://daily.dev/tags/neural-networks), [#ai-safety](https://daily.dev/tags/ai-safety) [View this post on daily.dev](https://daily.dev/posts/exposing-biases-moods-personalities-and-abstract-concepts-hidden-in-large-language-models-eyfxmvhsd) ```json {"@context":"https://schema.org","@graph":[{"@type":"Organization","@id":"https://daily.dev/#organization","name":"daily.dev","url":"https://daily.dev","logo":{"@type":"ImageObject","url":"https://daily.dev/apple-touch-icon.png","width":180,"height":180},"sameAs":["https://twitter.com/dailydotdev","https://github.com/dailydotdev","https://www.linkedin.com/company/daily-dev-ltd"]},{"@type":"WebSite","@id":"https://daily.dev/#website","url":"https://daily.dev","name":"daily.dev","publisher":{"@id":"https://daily.dev/#organization"},"potentialAction":{"@type":"SearchAction","target":{"@type":"EntryPoint","urlTemplate":"https://daily.dev/search?q={search_term_string}"},"query-input":"required name=search_term_string"}}]} {"@context":"https://schema.org","@type":"TechArticle","headline":"Exposing biases, moods, personalities, and abstract concepts hidden in large language models","url":"https://daily.dev/posts/exposing-biases-moods-personalities-and-abstract-concepts-hidden-in-large-language-models-eyfxmvhsd","mainEntityOfPage":{"@type":"WebPage","@id":"https://daily.dev/posts/exposing-biases-moods-personalities-and-abstract-concepts-hidden-in-large-language-models-eyfxmvhsd"},"datePublished":"2026-02-19T19:12:27.116Z","dateModified":"2026-02-19T19:19:52.557Z","description":"MIT and UC San Diego researchers have developed a method to detect and manipulate hidden abstract concepts—such as biases, moods, personalities, and...","image":"https://media.daily.dev/image/fetch/s--qn3-NjuK--/f_auto,pg_1,q_auto/https://news.mit.edu/sites/default/files/images/202602/MIT-LLM-Bias-01.gif?_a=BAMAMiZY0","thumbnailUrl":"https://media.daily.dev/image/fetch/s--qn3-NjuK--/f_auto,pg_1,q_auto/https://news.mit.edu/sites/default/files/images/202602/MIT-LLM-Bias-01.gif?_a=BAMAMiZY0","isAccessibleForFree":true,"articleSection":"MIT News","inLanguage":"en","publisher":{"@type":"Organization","name":"daily.dev","url":"https://daily.dev","logo":{"@type":"ImageObject","url":"https://daily.dev/apple-touch-icon.png","width":180,"height":180}},"author":{"@type":"Organization","name":"MIT News","logo":"https://media.daily.dev/image/upload/t_logo,f_auto/v1/logos/881e20a40aa44b12ba4bd27f5c14289d","url":"https://daily.dev/sources/mit"},"commentCount":0,"discussionUrl":"https://daily.dev/posts/exposing-biases-moods-personalities-and-abstract-concepts-hidden-in-large-language-models-eyfxmvhsd","interactionStatistic":[{"@type":"InteractionCounter","interactionType":{"@type":"LikeAction"},"userInteractionCount":0},{"@type":"InteractionCounter","interactionType":{"@type":"CommentAction"},"userInteractionCount":0}],"keywords":"llm,neural-networks,ai-safety","timeRequired":"PT7M"} {"@context":"https://schema.org","@type":"BreadcrumbList","itemListElement":[{"@type":"ListItem","position":1,"name":"Home","item":"https://daily.dev"},{"@type":"ListItem","position":2,"name":"MIT News","item":"https://daily.dev/sources/mit"},{"@type":"ListItem","position":3,"name":"Exposing biases, moods, personalities, and abstract concepts hidden in large language models"}]} ```