{"id":6908,"date":"2026-11-04T15:12:39","date_gmt":"2026-11-04T15:12:39","guid":{"rendered":"https:\/\/i10xnews.kinsta.cloud\/weight-space-programming-gpt-6-astra-qwen-edit"},"modified":"2026-11-04T15:12:39","modified_gmt":"2026-11-04T15:12:39","slug":"weight-space-programming-gpt-6-astra-qwen-edit","status":"publish","type":"post","link":"https:\/\/i10x.ai\/news\/weight-space-programming-gpt-6-astra-qwen-edit","title":{"rendered":"Weight-Space Programming Lets GPT-6 Astra Edit Qwen Weights"},"content":{"rendered":"\n<p class=\"wp-block-paragraph\">In a breakthrough that sidesteps the usual gradient descent process, researchers at MIT CSAIL have shown how one frontier model can directly read and rewrite the raw weights of another. OpenAI\u2019s GPT-6 Astra was used to adjust just 0.016% of a smaller Qwen model\u2019s parameters, lifting its MATH-500 score sixfold and pointing to a fresh way to debug &#8211; and perhaps improve &#8211; large language models on the fly.<\/p>\n\n\n\n<h2 class=\"wp-block-heading\">\u26a1 Quick Take<\/h2>\n\n\n\n<ul class=\"wp-block-list\"><li><strong>What happened:<\/strong> GPT-6 Astra directly read and edited 131,072 weights in Qwen3.5-0.8B without using a traditional training loop, optimizer, or gradient descent.<\/li><li><strong>The benchmark shock:<\/strong> The targeted edit spiked Qwen&#8217;s MATH-500 score from 7.9% to 48.1%, while also tripling its IFEval performance.<\/li><li><strong>The paradigm shift:<\/strong> Treating neural network weights as inspectable, editable code rather than black-box outputs of gradient descent opens the door to direct &#8220;model patching.&#8221;<\/li><li><strong>The hidden caveat:<\/strong> External evaluations suggest the edit didn&#8217;t instill raw mathematical genius from scratch; rather, it functioned as loop-repair, dropping the model&#8217;s answer truncation rate from 93% to roughly 33.1%.<\/li><\/ul>\n\n\n\n<h2 class=\"wp-block-heading\">What happened<\/h2>\n\n\n\n<p class=\"wp-block-paragraph\">Researchers at MIT CSAIL set out to see whether a capable frontier model could reach inside a smaller one and fix a concrete flaw. They asked OpenAI\u2019s GPT-6 Astra to examine the weights of Qwen3.5-0.8B and correct its habit of falling into repetitive loops on hard math problems.<\/p>\n\n\n\n<p class=\"wp-block-paragraph\">Rather than create fresh training data or run any fine-tuning step, Astra worked straight in weight space. It singled out 64 output columns in the MLP layers at positions 15 and 19, then changed exactly 131,072 weights &#8211; a mere 0.016% of the model\u2019s total. The effect was immediate. MATH-500 accuracy rose from 7.9% to 48.1%. External checks showed the change mainly stopped the model from looping until it hit its token limit, cutting the truncation rate from 93% to about 33.1%. In short, the patch removed a self-inflicted obstacle more than it taught new mathematics.<\/p>\n\n\n\n<h2 class=\"wp-block-heading\">Why it matters now<\/h2>\n\n\n\n<p class=\"wp-block-paragraph\">Weight-space programming undercuts the long-held view that the only reliable way to reshape a model is through heavy training runs. If another model can read, interpret, and rewrite weights directly, the field moves from training opaque \u201cbrains\u201d to editing readable code.<\/p>\n\n\n\n<p class=\"wp-block-paragraph\">For infrastructure teams this means the possibility of live model patching. Instead of curating datasets and retraining to fix hallucinations or safety issues, a frontier model could simply push a targeted diff. The compute savings could be substantial, and the path to recursive self-improvement becomes more concrete.<\/p>\n\n\n\n<p class=\"wp-block-paragraph\">At the same time the approach raises fresh safety questions. When weights can be altered outside any logged training run, traditional data-audit trails disappear. New cryptographic checks for model checkpoints will be needed to detect hidden changes.<\/p>\n\n\n\n<h2 class=\"wp-block-heading\">Who is most affected<\/h2>\n\n\n\n<figure class=\"wp-block-table\"><table class=\"has-fixed-layout\"><thead><tr><th>Stakeholder<\/th><th>Impact<\/th><th>Why<\/th><\/tr><\/thead><tbody><tr><td>ML Engineers &amp; Developers<\/td><td>Shift from fine-tuning to direct model patching.<\/td><td>Resolving specific failure modes (like reasoning loops) will require analyzing weight diffs rather than curating new RLHF datasets.<\/td><\/tr><tr><td>AI Safety &amp; Governance Teams<\/td><td>Nightmare auditability scenarios.<\/td><td>Models edited in weight-space bypass traditional training data audits, making hidden backdoors harder to detect without checkpoint verification.<\/td><\/tr><tr><td>Infrastructure Providers<\/td><td>Disruption to compute demand for alignment.<\/td><td>If smaller models can be aligned via surgical weight edits by an API-called frontier model, post-training compute requirements could drop.<\/td><\/tr><\/tbody><\/table><\/figure>\n\n\n\n<h2 class=\"wp-block-heading\">\u270d\ufe0f About the analysis<\/h2>\n\n\n\n<p class=\"wp-block-paragraph\">This is an independent i10x analysis built for CTOs, AI researchers, and technical founders tracking the evolution of AI infrastructure. It synthesizes current research claims, benchmark caveats, and external evaluations regarding MIT CSAIL&#8217;s weight-space programming experiments.<\/p>\n\n\n\n<ul class=\"wp-block-list\"><li><a href=\"https:\/\/medium.com\/@adnanmasood\/when-llms-edit-other-llms-weight-space-programming-and-the-conditions-for-recursive-2dceaec05403\" target=\"_blank\" rel=\"noreferrer noopener\">When LLMs Edit Other LLMs: Weight-Space Programming<\/a><\/li><li><a href=\"https:\/\/www.datacamp.com\/blog\/weight-space-programming\" target=\"_blank\" rel=\"noreferrer noopener\">Weight-Space Programming: How LLMs Read and Edit Neural Network Weights<\/a><\/li><li><a href=\"https:\/\/kingy.ai\/blog\/weight-space-programming-ai-agent-model-weights\/\" target=\"_blank\" rel=\"noreferrer noopener\">Weight-Space Programming: How GPT-6 Astra Raised Qwen\u2019s MATH-500 Score Sixfold<\/a><\/li><li><a href=\"https:\/\/www.theregister.com\/ai-and-ml\/2026\/11\/03\/forget-retraining-ai-can-now-tweak-another-ais-weights\/5307105\" target=\"_blank\" rel=\"noreferrer noopener\">Forget retraining &#8211; AI can now tweak another AI&#8217;s weights<\/a><\/li><li><a href=\"https:\/\/www.turingpost.com\/p\/what-is-weight-space-programming-and-why-it-suddenly-blew-up-the-internet\" target=\"_blank\" rel=\"noreferrer noopener\">FOD#174: What is Weight-Space Programming and why it suddenly blew up the internet?<\/a><\/li><li><a href=\"https:\/\/www.tokenpost.com\/news\/technology\/47287\" target=\"_blank\" rel=\"noreferrer noopener\">MIT Researchers Raise Qwen Math Score to 48.1% With Weight Edits<\/a><\/li><li><a href=\"https:\/\/www.systemdesign.academy\/explained\/weight-space-programming\" target=\"_blank\" rel=\"noreferrer noopener\">GPT-6 Astra edits Qwen&#8217;s weights: weight-space programming<\/a><\/li><li><a href=\"https:\/\/letsdatascience.com\/blog\/gpt-6-astra-edited-0-016-percent-qwen-weights-math-sixfold\" target=\"_blank\" rel=\"noreferrer noopener\">Weight-Space Programming: GPT-6 Astra Fixes Qwen&#8217;s Math | LDS<\/a><\/li><\/ul>\n\n\n\n<h2 class=\"wp-block-heading\">\ud83d\udd2d i10x Perspective<\/h2>\n\n\n\n<p class=\"wp-block-paragraph\">Weight-space programming shows that mechanistic interpretability has moved from theory to practical tool. Once weights are treated as an editable codebase rather than fixed results of gradient descent, the route to recursive self-improvement becomes clearer. We&#8217;ve seen the industry shift its focus from raw training scale to the intelligence needed to debug and rewrite models in real time; that shift now looks decisive.<\/p>\n","protected":false},"excerpt":{"rendered":"<p>Researchers at MIT CSAIL have shown that frontier models like GPT-6 Astra can directly read and rewrite the raw weights of smaller models such as Qwen3.5-0.8B. By adjusting just 0.016% of parameters without any gradient descent, the targeted edit boosted\u2026<\/p>\n","protected":false},"author":2,"featured_media":6907,"comment_status":"closed","ping_status":"closed","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":"9"},"categories":[9],"tags":[],"companies":[38],"products":[],"class_list":["post-6908","post","type-post","status-publish","format-standard","has-post-thumbnail","hentry","category-models","company-open-ai"],"yoast_head":"<!-- This site is optimized with the Yoast SEO plugin v28.7 - https:\/\/yoast.com\/product\/yoast-seo-wordpress\/ -->\n<title>Weight-Space Programming: GPT-6 Astra Edits Qwen Weights<\/title>\n<meta name=\"description\" content=\"MIT CSAIL researchers show GPT-6 Astra directly editing 0.016% of Qwen3.5-0.8B weights to raise MATH-500 score from 7.9% to 48.1%, bypassing gradient\u2026\" \/>\n<meta name=\"robots\" content=\"index, follow, max-snippet:-1, max-image-preview:large, max-video-preview:-1\" \/>\n<link rel=\"canonical\" href=\"https:\/\/i10xnews.kinsta.cloud\/weight-space-programming-gpt-6-astra-qwen-edit\" \/>\n<meta property=\"og:locale\" content=\"en_US\" \/>\n<meta property=\"og:type\" content=\"article\" \/>\n<meta property=\"og:title\" content=\"Weight-Space Programming: GPT-6 Astra Edits Qwen Weights\" \/>\n<meta property=\"og:description\" content=\"MIT CSAIL researchers show GPT-6 Astra directly editing 0.016% of Qwen3.5-0.8B weights to raise MATH-500 score from 7.9% to 48.1%, bypassing gradient\u2026\" \/>\n<meta property=\"og:url\" content=\"https:\/\/i10xnews.kinsta.cloud\/weight-space-programming-gpt-6-astra-qwen-edit\" \/>\n<meta property=\"og:site_name\" content=\"i10X News\" \/>\n<meta property=\"article:published_time\" content=\"2026-11-04T15:12:39+00:00\" \/>\n<meta property=\"og:image\" content=\"https:\/\/i10xnews.kinsta.cloud\/wp-content\/uploads\/2026\/11\/weight-space-programming-gpt-6-astra-qwen-edit-title.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=\"3 minutes\" \/>\n<script type=\"application\/ld+json\" class=\"yoast-schema-graph\">{\"@context\":\"https:\\\/\\\/schema.org\",\"@graph\":[{\"@type\":\"NewsArticle\",\"@id\":\"https:\\\/\\\/i10xnews.kinsta.cloud\\\/weight-space-programming-gpt-6-astra-qwen-edit#article\",\"isPartOf\":{\"@id\":\"https:\\\/\\\/i10xnews.kinsta.cloud\\\/weight-space-programming-gpt-6-astra-qwen-edit\"},\"author\":{\"name\":\"Christopher Ort\",\"@id\":\"https:\\\/\\\/i10x.ai\\\/news\\\/#\\\/schema\\\/person\\\/d410bb7ee72febefca2c9c369939219a\"},\"headline\":\"Weight-Space Programming Lets GPT-6 Astra Edit Qwen Weights\",\"datePublished\":\"2026-11-04T15:12:39+00:00\",\"mainEntityOfPage\":{\"@id\":\"https:\\\/\\\/i10xnews.kinsta.cloud\\\/weight-space-programming-gpt-6-astra-qwen-edit\"},\"wordCount\":725,\"publisher\":{\"@id\":\"https:\\\/\\\/i10x.ai\\\/news\\\/#organization\"},\"image\":{\"@id\":\"https:\\\/\\\/i10xnews.kinsta.cloud\\\/weight-space-programming-gpt-6-astra-qwen-edit#primaryimage\"},\"thumbnailUrl\":\"https:\\\/\\\/i10x.ai\\\/news\\\/wp-content\\\/uploads\\\/2026\\\/11\\\/weight-space-programming-gpt-6-astra-qwen-edit-title.webp\",\"articleSection\":[\"Models\"],\"inLanguage\":\"en-US\",\"dateModified\":\"2026-11-04T15:12:39+00:00\"},{\"@type\":\"WebPage\",\"@id\":\"https:\\\/\\\/i10xnews.kinsta.cloud\\\/weight-space-programming-gpt-6-astra-qwen-edit\",\"url\":\"https:\\\/\\\/i10xnews.kinsta.cloud\\\/weight-space-programming-gpt-6-astra-qwen-edit\",\"name\":\"Weight-Space Programming: GPT-6 Astra Edits Qwen Weights\",\"isPartOf\":{\"@id\":\"https:\\\/\\\/i10x.ai\\\/news\\\/#website\"},\"primaryImageOfPage\":{\"@id\":\"https:\\\/\\\/i10xnews.kinsta.cloud\\\/weight-space-programming-gpt-6-astra-qwen-edit#primaryimage\"},\"image\":{\"@id\":\"https:\\\/\\\/i10xnews.kinsta.cloud\\\/weight-space-programming-gpt-6-astra-qwen-edit#primaryimage\"},\"thumbnailUrl\":\"https:\\\/\\\/i10x.ai\\\/news\\\/wp-content\\\/uploads\\\/2026\\\/11\\\/weight-space-programming-gpt-6-astra-qwen-edit-title.webp\",\"datePublished\":\"2026-11-04T15:12:39+00:00\",\"description\":\"MIT CSAIL researchers show GPT-6 Astra directly editing 0.016% of Qwen3.5-0.8B weights to raise MATH-500 score from 7.9% to 48.1%, bypassing gradient\u2026\",\"breadcrumb\":{\"@id\":\"https:\\\/\\\/i10xnews.kinsta.cloud\\\/weight-space-programming-gpt-6-astra-qwen-edit#breadcrumb\"},\"inLanguage\":\"en-US\",\"potentialAction\":[{\"@type\":\"ReadAction\",\"target\":[\"https:\\\/\\\/i10xnews.kinsta.cloud\\\/weight-space-programming-gpt-6-astra-qwen-edit\"]}]},{\"@type\":\"ImageObject\",\"inLanguage\":\"en-US\",\"@id\":\"https:\\\/\\\/i10xnews.kinsta.cloud\\\/weight-space-programming-gpt-6-astra-qwen-edit#primaryimage\",\"url\":\"https:\\\/\\\/i10x.ai\\\/news\\\/wp-content\\\/uploads\\\/2026\\\/11\\\/weight-space-programming-gpt-6-astra-qwen-edit-title.webp\",\"contentUrl\":\"https:\\\/\\\/i10x.ai\\\/news\\\/wp-content\\\/uploads\\\/2026\\\/11\\\/weight-space-programming-gpt-6-astra-qwen-edit-title.webp\",\"width\":1344,\"height\":768,\"caption\":\"Illustration: i10X\"},{\"@type\":\"BreadcrumbList\",\"@id\":\"https:\\\/\\\/i10xnews.kinsta.cloud\\\/weight-space-programming-gpt-6-astra-qwen-edit#breadcrumb\",\"itemListElement\":[{\"@type\":\"ListItem\",\"position\":1,\"name\":\"Home\",\"item\":\"https:\\\/\\\/i10x.ai\\\/news\"},{\"@type\":\"ListItem\",\"position\":2,\"name\":\"Weight-Space Programming Lets GPT-6 Astra Edit Qwen Weights\"}]},{\"@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":"Weight-Space Programming: GPT-6 Astra Edits Qwen Weights","description":"MIT CSAIL researchers show GPT-6 Astra directly editing 0.016% of Qwen3.5-0.8B weights to raise MATH-500 score from 7.9% to 48.1%, bypassing gradient\u2026","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:\/\/i10xnews.kinsta.cloud\/weight-space-programming-gpt-6-astra-qwen-edit","og_locale":"en_US","og_type":"article","og_title":"Weight-Space Programming: GPT-6 Astra Edits Qwen Weights","og_description":"MIT CSAIL researchers show GPT-6 Astra directly editing 0.016% of Qwen3.5-0.8B weights to raise MATH-500 score from 7.9% to 48.1%, bypassing gradient\u2026","og_url":"https:\/\/i10xnews.kinsta.cloud\/weight-space-programming-gpt-6-astra-qwen-edit","og_site_name":"i10X News","article_published_time":"2026-11-04T15:12:39+00:00","og_image":[{"width":1344,"height":768,"url":"https:\/\/i10xnews.kinsta.cloud\/wp-content\/uploads\/2026\/11\/weight-space-programming-gpt-6-astra-qwen-edit-title.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":"3 minutes"},"schema":{"@context":"https:\/\/schema.org","@graph":[{"@type":"NewsArticle","@id":"https:\/\/i10xnews.kinsta.cloud\/weight-space-programming-gpt-6-astra-qwen-edit#article","isPartOf":{"@id":"https:\/\/i10xnews.kinsta.cloud\/weight-space-programming-gpt-6-astra-qwen-edit"},"author":{"name":"Christopher Ort","@id":"https:\/\/i10x.ai\/news\/#\/schema\/person\/d410bb7ee72febefca2c9c369939219a"},"headline":"Weight-Space Programming Lets GPT-6 Astra Edit Qwen Weights","datePublished":"2026-11-04T15:12:39+00:00","mainEntityOfPage":{"@id":"https:\/\/i10xnews.kinsta.cloud\/weight-space-programming-gpt-6-astra-qwen-edit"},"wordCount":725,"publisher":{"@id":"https:\/\/i10x.ai\/news\/#organization"},"image":{"@id":"https:\/\/i10xnews.kinsta.cloud\/weight-space-programming-gpt-6-astra-qwen-edit#primaryimage"},"thumbnailUrl":"https:\/\/i10x.ai\/news\/wp-content\/uploads\/2026\/11\/weight-space-programming-gpt-6-astra-qwen-edit-title.webp","articleSection":["Models"],"inLanguage":"en-US","dateModified":"2026-11-04T15:12:39+00:00"},{"@type":"WebPage","@id":"https:\/\/i10xnews.kinsta.cloud\/weight-space-programming-gpt-6-astra-qwen-edit","url":"https:\/\/i10xnews.kinsta.cloud\/weight-space-programming-gpt-6-astra-qwen-edit","name":"Weight-Space Programming: GPT-6 Astra Edits Qwen Weights","isPartOf":{"@id":"https:\/\/i10x.ai\/news\/#website"},"primaryImageOfPage":{"@id":"https:\/\/i10xnews.kinsta.cloud\/weight-space-programming-gpt-6-astra-qwen-edit#primaryimage"},"image":{"@id":"https:\/\/i10xnews.kinsta.cloud\/weight-space-programming-gpt-6-astra-qwen-edit#primaryimage"},"thumbnailUrl":"https:\/\/i10x.ai\/news\/wp-content\/uploads\/2026\/11\/weight-space-programming-gpt-6-astra-qwen-edit-title.webp","datePublished":"2026-11-04T15:12:39+00:00","description":"MIT CSAIL researchers show GPT-6 Astra directly editing 0.016% of Qwen3.5-0.8B weights to raise MATH-500 score from 7.9% to 48.1%, bypassing gradient\u2026","breadcrumb":{"@id":"https:\/\/i10xnews.kinsta.cloud\/weight-space-programming-gpt-6-astra-qwen-edit#breadcrumb"},"inLanguage":"en-US","potentialAction":[{"@type":"ReadAction","target":["https:\/\/i10xnews.kinsta.cloud\/weight-space-programming-gpt-6-astra-qwen-edit"]}]},{"@type":"ImageObject","inLanguage":"en-US","@id":"https:\/\/i10xnews.kinsta.cloud\/weight-space-programming-gpt-6-astra-qwen-edit#primaryimage","url":"https:\/\/i10x.ai\/news\/wp-content\/uploads\/2026\/11\/weight-space-programming-gpt-6-astra-qwen-edit-title.webp","contentUrl":"https:\/\/i10x.ai\/news\/wp-content\/uploads\/2026\/11\/weight-space-programming-gpt-6-astra-qwen-edit-title.webp","width":1344,"height":768,"caption":"Illustration: i10X"},{"@type":"BreadcrumbList","@id":"https:\/\/i10xnews.kinsta.cloud\/weight-space-programming-gpt-6-astra-qwen-edit#breadcrumb","itemListElement":[{"@type":"ListItem","position":1,"name":"Home","item":"https:\/\/i10x.ai\/news"},{"@type":"ListItem","position":2,"name":"Weight-Space Programming Lets GPT-6 Astra Edit Qwen Weights"}]},{"@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\/6908","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=6908"}],"version-history":[{"count":0,"href":"https:\/\/i10x.ai\/news\/wp-json\/wp\/v2\/posts\/6908\/revisions"}],"wp:featuredmedia":[{"embeddable":true,"href":"https:\/\/i10x.ai\/news\/wp-json\/wp\/v2\/media\/6907"}],"wp:attachment":[{"href":"https:\/\/i10x.ai\/news\/wp-json\/wp\/v2\/media?parent=6908"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/i10x.ai\/news\/wp-json\/wp\/v2\/categories?post=6908"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/i10x.ai\/news\/wp-json\/wp\/v2\/tags?post=6908"},{"taxonomy":"company","embeddable":true,"href":"https:\/\/i10x.ai\/news\/wp-json\/wp\/v2\/companies?post=6908"},{"taxonomy":"product","embeddable":true,"href":"https:\/\/i10x.ai\/news\/wp-json\/wp\/v2\/products?post=6908"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}