--- title: AI-assisted learning platform helps robots reason beyond... description: MIT and NYU researchers built FLEX, a pipeline that lets robots generalize beyond literal natural-language commands during motion planning. Told 'don't spill... canonical: https://daily.dev/posts/ai-assisted-learning-platform-helps-robots-reason-beyond-literal-commands-8dgcjrslz twitter:card: summary_large_image twitter:site: @dailydotdev og:type: website og:site_name: daily.dev og:title: AI-assisted learning platform helps robots reason beyond literal commands | daily.dev og:description: MIT and NYU researchers built FLEX, a pipeline that lets robots generalize beyond literal natural-language commands during motion planning. Told 'don't spill... og:url: https://daily.dev/posts/ai-assisted-learning-platform-helps-robots-reason-beyond-literal-commands-8dgcjrslz og:image: https://api.daily.dev/og/posts/8DGcJRSlZ.png og:image:alt: AI-assisted learning platform helps robots reason beyond literal commands 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. # AI-assisted learning platform helps robots reason beyond literal commands **[MIT News](https://daily.dev/sources/mit)** · 6 min read · 0 upvotes · 0 comments ## Summary MIT and NYU researchers built FLEX, a pipeline that lets robots generalize beyond literal natural-language commands during motion planning. Told 'don't spill on my laptop,' a robot using FLEX also avoids electronics like a nearby phone. FLEX works in three steps: Add, which lists a stated object to avoid; Generalize, which uses GPT-4o to make a generalization that fragile objects should be avoided; and Mask, which removes objects no longer present using computer vision. The whole process takes about five seconds and was tested in simulations involving coffee, a candle, and a bowling ball. ## Full article daily.dev links to this article rather than hosting it. Read it at the original source: ## Questions this post answers ### How does the FLEX robot learning platform use GPT-4o to generalize safety instructions? FLEX uses GPT-4o in its 'Generalize' step to find patterns among objects a person has told the robot to avoid, then infers a shared property like fragility so the robot avoids similar unstated objects too. For example, after being told not to spill coffee on a laptop, the robot also avoids a nearby phone without being told explicitly. _Developers exploring LLM reasoning beyond text can follow applied robotics work like this on daily.dev._ ### What are the three steps in MIT's FLEX robot motion-planning pipeline? FLEX works in three steps: Add, where the robot adds a user-specified object to its list of things to avoid; Generalize, where an LLM like GPT-4o infers related objects sharing a property such as fragility; and Mask, where computer vision removes objects no longer present in the environment. The whole process takes about five seconds. _Engineers combining LLMs with robotics pipelines can track approaches like this on daily.dev._ ## Similar posts on daily.dev - [A neural blueprint for human-like intelligence in soft robots](https://daily.dev/posts/a-neural-blueprint-for-human-like-intelligence-in-soft-robots-7efw2mkgi) · MIT News · 0 upvotes · 0 comments - [AI agents create virtual playgrounds to help robots get crucial training data](https://daily.dev/posts/ai-agents-create-virtual-playgrounds-to-help-robots-get-crucial-training-data-paryx7xxz) · Robohub · 0 upvotes · 0 comments - [Smarter Robot Emotions From Vision Language Models](https://daily.dev/posts/smarter-robot-emotions-from-vision-language-models-sjiof9n2a) · IEEE Spectrum · 1 upvotes · 0 comments --- Tags: [#llm](https://daily.dev/tags/llm), [#nlp](https://daily.dev/tags/nlp), [#robotics](https://daily.dev/tags/robotics), [#mit](https://daily.dev/tags/mit) [View this post on daily.dev](https://daily.dev/posts/ai-assisted-learning-platform-helps-robots-reason-beyond-literal-commands-8dgcjrslz) ```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":"AI-assisted learning platform helps robots reason beyond literal commands","url":"https://daily.dev/posts/ai-assisted-learning-platform-helps-robots-reason-beyond-literal-commands-8dgcjrslz","mainEntityOfPage":{"@type":"WebPage","@id":"https://daily.dev/posts/ai-assisted-learning-platform-helps-robots-reason-beyond-literal-commands-8dgcjrslz"},"datePublished":"2026-11-09T05:04:54.840Z","dateModified":"2026-11-09T05:05:22.390Z","description":"MIT and NYU researchers built FLEX, a pipeline that lets robots generalize beyond literal natural-language commands during motion planning. Told 'don't spill...","image":"https://media.daily.dev/image/upload/f_auto,q_auto/v1/posts/55e3d26e1d73599ef0d110b571ce27f7?_a=AQAEuop","thumbnailUrl":"https://media.daily.dev/image/upload/f_auto,q_auto/v1/posts/55e3d26e1d73599ef0d110b571ce27f7?_a=AQAEuop","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/ai-assisted-learning-platform-helps-robots-reason-beyond-literal-commands-8dgcjrslz","interactionStatistic":[{"@type":"InteractionCounter","interactionType":{"@type":"LikeAction"},"userInteractionCount":0},{"@type":"InteractionCounter","interactionType":{"@type":"CommentAction"},"userInteractionCount":0}],"keywords":"llm,nlp,robotics,mit","timeRequired":"PT6M"} {"@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":"AI-assisted learning platform helps robots reason beyond literal commands"}]} {"@context":"https://schema.org","@type":"FAQPage","@id":"https://daily.dev/posts/ai-assisted-learning-platform-helps-robots-reason-beyond-literal-commands-8dgcjrslz#faq","mainEntity":[{"@type":"Question","name":"How does the FLEX robot learning platform use GPT-4o to generalize safety instructions?","acceptedAnswer":{"@type":"Answer","text":"FLEX uses GPT-4o in its 'Generalize' step to find patterns among objects a person has told the robot to avoid, then infers a shared property like fragility so the robot avoids similar unstated objects too. For example, after being told not to spill coffee on a laptop, the robot also avoids a nearby phone without being told explicitly. Developers exploring LLM reasoning beyond text can follow applied robotics work like this on daily.dev."}},{"@type":"Question","name":"What are the three steps in MIT's FLEX robot motion-planning pipeline?","acceptedAnswer":{"@type":"Answer","text":"FLEX works in three steps: Add, where the robot adds a user-specified object to its list of things to avoid; Generalize, where an LLM like GPT-4o infers related objects sharing a property such as fragility; and Mask, where computer vision removes objects no longer present in the environment. The whole process takes about five seconds. Engineers combining LLMs with robotics pipelines can track approaches like this on daily.dev."}}]} ```