--- title: A better way to turn 2D designs into 3D models for rapid... description: MIT researchers developed GIFT (Geometric Inference Feedback Tuning), a system that trains vision-language models to convert 2D designs into accurate CAD... canonical: https://daily.dev/posts/a-better-way-to-turn-2d-designs-into-3d-models-for-rapid-prototyping-2smhu1y25 twitter:card: summary_large_image twitter:site: @dailydotdev og:type: website og:site_name: daily.dev og:title: A better way to turn 2D designs into 3D models for rapid prototyping | daily.dev og:description: MIT researchers developed GIFT (Geometric Inference Feedback Tuning), a system that trains vision-language models to convert 2D designs into accurate CAD... og:url: https://daily.dev/posts/a-better-way-to-turn-2d-designs-into-3d-models-for-rapid-prototyping-2smhu1y25 og:image: https://api.daily.dev/og/posts/2Smhu1y25.png og:image:alt: A better way to turn 2D designs into 3D models for rapid prototyping 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. # A better way to turn 2D designs into 3D models for rapid prototyping **[MIT News](https://daily.dev/sources/mit)** · 6 min read · 1 upvotes · 0 comments ## Summary MIT researchers developed GIFT (Geometric Inference Feedback Tuning), a system that trains vision-language models to convert 2D designs into accurate CAD programs for 3D modeling. Rather than relying on traditional random data augmentation, GIFT tests the model repeatedly on CAD generation tasks, identifies near-miss failures, corrects them, and incorporates both failures and successes into a new training dataset. This model-aware and task-aware approach uses inference-time scaling to improve a pre-trained model without full retraining, requiring only about 20% of the computation of competing methods while producing more geometrically accurate 3D models. The technique could accelerate rapid prototyping workflows in engineering and help surface design choices engineers might otherwise miss. ## Full article daily.dev links to this article rather than hosting it. Read it at the original source: ## Questions this post answers ### What is GIFT in the context of AI-generated CAD models? GIFT (Geometric Inference Feedback Tuning) is a data augmentation framework developed by MIT and collaborators that teaches vision-language models to convert 2D images into accurate CAD programs. It tests a model on CAD generation tasks, captures near-miss and successful solutions, and builds a targeted dataset from those results, improving accuracy while using about 20 percent as much computation as competing methods. _Engineers exploring AI-assisted CAD workflows can follow research like this via daily.dev._ ### How does model-aware data augmentation differ from traditional data augmentation? Traditional data augmentation randomly tweaks existing data, such as adjusting color, size, or shape in images, to create more training samples. Model-aware augmentation, as used in GIFT, instead tests a specific model to find its strengths and weaknesses, then generates new data targeted at the problems it struggles with, particularly cases solved correctly only part of the time, such as 50 percent. _daily.dev helps AI practitioners track emerging techniques like model-aware data augmentation._ ## Similar posts on daily.dev - [AI Just Rewired 3D Modeling. Here’s What It Means for Every Designer Holding a Mouse Right Now](https://daily.dev/posts/ai-just-rewired-3d-modeling-here-s-what-it-means-for-every-designer-holding-a-mouse-right-now-wyt4gcqmo) · Medium · 0 upvotes · 0 comments - [New AI agent learns to use CAD to create 3D objects from sketches](https://daily.dev/posts/new-ai-agent-learns-to-use-cad-to-create-3d-objects-from-sketches-kj6uoxx6u) · MIT News · 0 upvotes · 0 comments - [“Robot, make me a chair”](https://daily.dev/posts/robot-make-me-a-chair--equjakger) · Robohub · 0 upvotes · 0 comments - [GenCAD](https://daily.dev/posts/gencad-issaibdx4) · Hacker News · 0 upvotes · 0 comments --- Tags: [#data-science](https://daily.dev/tags/data-science), [#genai](https://daily.dev/tags/genai), [#vlm](https://daily.dev/tags/vlm) [View this post on daily.dev](https://daily.dev/posts/a-better-way-to-turn-2d-designs-into-3d-models-for-rapid-prototyping-2smhu1y25) ```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":"A better way to turn 2D designs into 3D models for rapid prototyping","url":"https://daily.dev/posts/a-better-way-to-turn-2d-designs-into-3d-models-for-rapid-prototyping-2smhu1y25","mainEntityOfPage":{"@type":"WebPage","@id":"https://daily.dev/posts/a-better-way-to-turn-2d-designs-into-3d-models-for-rapid-prototyping-2smhu1y25"},"datePublished":"2026-07-16T04:01:42.694Z","dateModified":"2026-09-13T19:52:14.154Z","description":"MIT researchers developed GIFT (Geometric Inference Feedback Tuning), a system that trains vision-language models to convert 2D designs into accurate CAD...","image":"https://media.daily.dev/image/upload/f_auto,q_auto/v1/posts/800d9cbfb439cc53843f834fe803f885?_a=AQAEuop","thumbnailUrl":"https://media.daily.dev/image/upload/f_auto,q_auto/v1/posts/800d9cbfb439cc53843f834fe803f885?_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/a-better-way-to-turn-2d-designs-into-3d-models-for-rapid-prototyping-2smhu1y25","interactionStatistic":[{"@type":"InteractionCounter","interactionType":{"@type":"LikeAction"},"userInteractionCount":1},{"@type":"InteractionCounter","interactionType":{"@type":"CommentAction"},"userInteractionCount":0}],"keywords":"data-science,genai,vlm","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":"A better way to turn 2D designs into 3D models for rapid prototyping"}]} {"@context":"https://schema.org","@type":"FAQPage","@id":"https://daily.dev/posts/a-better-way-to-turn-2d-designs-into-3d-models-for-rapid-prototyping-2smhu1y25#faq","mainEntity":[{"@type":"Question","name":"What is GIFT in the context of AI-generated CAD models?","acceptedAnswer":{"@type":"Answer","text":"GIFT (Geometric Inference Feedback Tuning) is a data augmentation framework developed by MIT and collaborators that teaches vision-language models to convert 2D images into accurate CAD programs. It tests a model on CAD generation tasks, captures near-miss and successful solutions, and builds a targeted dataset from those results, improving accuracy while using about 20 percent as much computation as competing methods. Engineers exploring AI-assisted CAD workflows can follow research like this via daily.dev."}},{"@type":"Question","name":"How does model-aware data augmentation differ from traditional data augmentation?","acceptedAnswer":{"@type":"Answer","text":"Traditional data augmentation randomly tweaks existing data, such as adjusting color, size, or shape in images, to create more training samples. Model-aware augmentation, as used in GIFT, instead tests a specific model to find its strengths and weaknesses, then generates new data targeted at the problems it struggles with, particularly cases solved correctly only part of the time, such as 50 percent. daily.dev helps AI practitioners track emerging techniques like model-aware data augmentation."}}]} ```