{"id":6138,"date":"2026-10-19T16:49:19","date_gmt":"2026-10-19T16:49:19","guid":{"rendered":"https:\/\/i10xnews.kinsta.cloud\/?p=6138"},"modified":"2026-10-19T17:20:09","modified_gmt":"2026-10-19T17:20:09","slug":"ride-rl-induced-direction-extrapolation-distillation","status":"publish","type":"post","link":"https:\/\/i10x.ai\/news\/ride-rl-induced-direction-extrapolation-distillation","title":{"rendered":"Distill the RL Direction in Hidden Space, Not Noisy Output Rewards"},"content":{"rendered":"\n<h2 class=\"wp-block-heading\">\u26a1 Quick Take<\/h2>\n\n<ul class=\"wp-block-list\"><li><strong>The finding:<\/strong> On-policy distillation that extrapolates an implicit output-space reward is unstable because the LM head attenuates representation changes anisotropically.<\/li><li><strong>The fix:<\/strong> RIDE extrapolates the RL-induced residual in hidden space at every layer. It improves over OPRD on every pair and over ExOPD by 9.1 points on average.<\/li><li><strong>The smoking gun:<\/strong> The head&#8217;s 512 weakest directions carry 79.8% of the teacher-to-base residual&#8217;s hidden energy but only 30.0% of its centered-logit energy.<\/li><\/ul>\n\n\n<p class=\"wp-block-paragraph\">On-policy distillation trains a student on its own trajectories to match a teacher. Generalized variants try to surpass the teacher by extrapolating an implicit reward in output space \u2014 and often become unstable. Researchers from USTC, Rutgers, Zhejiang University and collaborators propose doing the extrapolation where RL actually moved the model: in representations.<\/p>\n\n\n\n<figure class=\"wp-block-image size-large\"><img loading=\"lazy\" decoding=\"async\" width=\"1344\" height=\"768\" src=\"https:\/\/i10xnews.kinsta.cloud\/wp-content\/uploads\/2026\/10\/ride-rl-induced-direction-extrapolation-distillation-head.webp\" alt=\"Stat card: weakest head directions hold 79.8 percent of hidden residual energy but only 30 percent of logit energy; RIDE beats ExOPD by 9.1 points on average.\" class=\"wp-image-6137\" srcset=\"https:\/\/i10x.ai\/news\/wp-content\/uploads\/2026\/10\/ride-rl-induced-direction-extrapolation-distillation-head.webp 1344w, https:\/\/i10x.ai\/news\/wp-content\/uploads\/2026\/10\/ride-rl-induced-direction-extrapolation-distillation-head-300x171.webp 300w, https:\/\/i10x.ai\/news\/wp-content\/uploads\/2026\/10\/ride-rl-induced-direction-extrapolation-distillation-head-1024x585.webp 1024w, https:\/\/i10x.ai\/news\/wp-content\/uploads\/2026\/10\/ride-rl-induced-direction-extrapolation-distillation-head-768x439.webp 768w\" sizes=\"auto, (max-width: 1344px) 100vw, 1344px\" \/><figcaption class=\"wp-element-caption\">i10X chart based on data from Li et al. (2026), arXiv:2609.36484, Figure 1 and main results (arXiv non-exclusive license).<\/figcaption><\/figure>\n\n\n\n<h2 class=\"wp-block-heading\">\ud83d\udd2c What the study shows<\/h2>\n\n\n<p class=\"wp-block-paragraph\">Across four base\/RL-teacher pairs, ExOPD can fall below its teacher and even below untouched students when the teacher moved little. RIDE steadily improves for \u03bb in 0.75\u20131.35 with format scores of 94\u201397%.<\/p>\n\n\n<ul class=\"wp-block-list\"><li><strong>Margin over OPRD isolates the idea:<\/strong> Because RIDE shares OPRD&#8217;s protocol and differs only in target, its 0.97\u20134.06 point gains attribute to residual extrapolation.<\/li><li><strong>ExOPD fails when RL moved a lot or a little:<\/strong> On an R1-Distill pair where RL moved the teacher 16.3 points, ExOPD underperforms output-space matching baselines; on small-move pairs it can fall 14.1 points below the student.<\/li><\/ul>\n\n\n<h2 class=\"wp-block-heading\">\u2705 Why it matters: distill where the update lives<\/h2>\n\n\n<p class=\"wp-block-paragraph\">If RL rewires mid-layers but the head squashes that signal, matching logits alone under-supervises the student.<\/p>\n\n\n\n<h2 class=\"wp-block-heading\">\ud83e\uddea Method in brief<\/h2>\n\n\n<p class=\"wp-block-paragraph\">Measure RL-induced direction vs base at every layer; extrapolate with coefficient \u03bb; compare Avg@16 against OPD, OPRD and ExOPD.<\/p>\n\n\n\n<h2 class=\"wp-block-heading\">\u2753 What remains open<\/h2>\n\n<ul class=\"wp-block-list\"><li><strong>Teacher must be RL-shifted:<\/strong> Method assumes a measurable residual from a base checkpoint.<\/li><li><strong>Hyperparameter \u03bb:<\/strong> Too large \u03bb eventually hurts ExOPD-style runs; RIDE&#8217;s safe band is reported but still tuned.<\/li><\/ul>\n\n\n<h2 class=\"wp-block-heading\">\ud83e\udded Context<\/h2>\n\n\n<p class=\"wp-block-paragraph\">Related distillation work in <a href=\"\/news\/category\/studies\/reasoning\">Reasoning<\/a> and <a href=\"\/news\/category\/studies\/efficiency-training\">Efficiency &amp; Training<\/a>.<\/p>\n\n\n\n<h2 class=\"wp-block-heading\">\ud83d\udee0\ufe0f What it means for distillation stacks<\/h2>\n\n<ul class=\"wp-block-list\"><li><strong>Keep base checkpoints:<\/strong> You need them to estimate the RL direction.<\/li><li><strong>Prefer representation targets:<\/strong> When extrapolating past the teacher.<\/li><li><strong>Monitor format collapse:<\/strong> A sign \u03bb is too aggressive.<\/li><\/ul>\n\n\n<h2 class=\"wp-block-heading\">\u270d\ufe0f About the analysis<\/h2>\n\n\n<p class=\"wp-block-paragraph\">This article is based on the full text of the preprint &#8220;The Teacher Is a Direction, Not a Destination: Extrapolating RL-Induced Representation Residuals in On-Policy Distillation&#8221; (arXiv:2609.36484v2, submitted September 29, 2026). All numbers are taken from the paper&#8217;s text, tables and figures; i10X did not rerun the experiments. The paper is distributed under arXiv&#8217;s non-exclusive license, so i10X does not reproduce its figures; the chart was drawn by i10X from numbers reported in Figure 1 and main results.<\/p>\n\n\n\n<h2 class=\"wp-block-heading\">\ud83d\udd2d i10x Perspective<\/h2>\n\n\n<p class=\"wp-block-paragraph\">RIDE\u2019s metaphor is right: the teacher is a direction, not a destination. Extrapolating that direction in the space where RL left fingerprints is a cleaner idea than amplifying noisy token-level reward ratios.<\/p>\n\n","protected":false},"excerpt":{"rendered":"<p>RIDE extrapolates RL-induced representation shifts layer-wise. It beats output-space ExOPD by 9.1 points on average and improves over OPRD by 0.97\u20134.06 points across four base\/teacher pairs.<\/p>\n","protected":false},"author":2,"featured_media":6288,"comment_status":"open","ping_status":"open","sticky":false,"template":"","format":"standard","meta":{"paper_title":"The Teacher Is a Direction, Not a Destination: Extrapolating RL-Induced Representation Residuals in On-Policy Distillation","paper_url":"https:\/\/arxiv.org\/abs\/2609.36484","pdf_url":"https:\/\/arxiv.org\/pdf\/2609.36484v2","doi":"","arxiv_id":"2609.36484","authors":"Hao Li, MeiJia Chen, Weijie Ren, Donghan Li, Zijun Tian, Jingchun Huang, Naibo Wang","institutions":"University of Science and Technology of China; Rutgers University; Zhejiang University; independent researchers","paper_date":"2026-09-29","venue":"arXiv (cs.LG), version 2","peer_reviewed":false,"code_url":"","data_url":"","funding_coi":"See paper.","figure_license":"Chart drawn by i10X from Figure 1 and main results (arXiv non-exclusive license)","paper_license":"arXiv non-exclusive license","replication":"no","study_models":"Four base\/RL-teacher pairs incl. Qwen3-4B and R1-Distill lineages","study_sample":"Avg@16 evaluations with prompt-bootstrap CIs","footnotes":"","_yoast_wpseo_primary_category":""},"categories":[56,52],"tags":[],"companies":[],"products":[],"class_list":["post-6138","post","type-post","status-publish","format-standard","has-post-thumbnail","hentry","category-reasoning","category-studies"],"yoast_head":"<!-- This site is optimized with the Yoast SEO plugin v28.7 - https:\/\/yoast.com\/product\/yoast-seo-wordpress\/ -->\n<title>RIDE: Extrapolating RL Residuals for Better Distillation - i10X News<\/title>\n<meta name=\"description\" content=\"RIDE distills RL teachers via representation residuals. +9.1 vs ExOPD on average; more stable than output-space extrapolation.\" \/>\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\/ride-rl-induced-direction-extrapolation-distillation\" \/>\n<meta property=\"og:locale\" content=\"en_US\" \/>\n<meta property=\"og:type\" content=\"article\" \/>\n<meta property=\"og:title\" content=\"RIDE: Extrapolating RL Residuals for Better Distillation - i10X News\" \/>\n<meta property=\"og:description\" content=\"RIDE distills RL teachers via representation residuals. +9.1 vs ExOPD on average; more stable than output-space extrapolation.\" \/>\n<meta property=\"og:url\" content=\"https:\/\/i10x.ai\/news\/ride-rl-induced-direction-extrapolation-distillation\" \/>\n<meta property=\"og:site_name\" content=\"i10X News\" \/>\n<meta property=\"article:published_time\" content=\"2026-10-19T16:49:19+00:00\" \/>\n<meta property=\"article:modified_time\" content=\"2026-10-19T17:20:09+00:00\" \/>\n<meta property=\"og:image\" content=\"https:\/\/i10x.ai\/news\/wp-content\/uploads\/2026\/10\/ride-rl-induced-direction-extrapolation-distillation-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=\"2 minutes\" \/>\n<script type=\"application\/ld+json\" class=\"yoast-schema-graph\">{\"@context\":\"https:\\\/\\\/schema.org\",\"@graph\":[{\"@type\":\"NewsArticle\",\"@id\":\"https:\\\/\\\/i10x.ai\\\/news\\\/ride-rl-induced-direction-extrapolation-distillation#article\",\"isPartOf\":{\"@id\":\"https:\\\/\\\/i10x.ai\\\/news\\\/ride-rl-induced-direction-extrapolation-distillation\"},\"author\":{\"name\":\"Christopher Ort\",\"@id\":\"https:\\\/\\\/news.i10x.ai\\\/#\\\/schema\\\/person\\\/d410bb7ee72febefca2c9c369939219a\"},\"headline\":\"Distill the RL Direction in Hidden Space, Not Noisy Output Rewards\",\"datePublished\":\"2026-10-19T16:49:19+00:00\",\"dateModified\":\"2026-10-19T17:20:09+00:00\",\"mainEntityOfPage\":{\"@id\":\"https:\\\/\\\/i10x.ai\\\/news\\\/ride-rl-induced-direction-extrapolation-distillation\"},\"wordCount\":486,\"commentCount\":0,\"publisher\":{\"@id\":\"https:\\\/\\\/news.i10x.ai\\\/#organization\"},\"image\":{\"@id\":\"https:\\\/\\\/i10x.ai\\\/news\\\/ride-rl-induced-direction-extrapolation-distillation#primaryimage\"},\"thumbnailUrl\":\"https:\\\/\\\/i10x.ai\\\/news\\\/wp-content\\\/uploads\\\/2026\\\/10\\\/ride-rl-induced-direction-extrapolation-distillation-title.webp\",\"articleSection\":[\"Reasoning\",\"Studies\"],\"inLanguage\":\"en-US\",\"potentialAction\":[{\"@type\":\"CommentAction\",\"name\":\"Comment\",\"target\":[\"https:\\\/\\\/i10x.ai\\\/news\\\/ride-rl-induced-direction-extrapolation-distillation#respond\"]}]},{\"@type\":\"WebPage\",\"@id\":\"https:\\\/\\\/i10x.ai\\\/news\\\/ride-rl-induced-direction-extrapolation-distillation\",\"url\":\"https:\\\/\\\/i10x.ai\\\/news\\\/ride-rl-induced-direction-extrapolation-distillation\",\"name\":\"RIDE: Extrapolating RL Residuals for Better Distillation - i10X News\",\"isPartOf\":{\"@id\":\"https:\\\/\\\/news.i10x.ai\\\/#website\"},\"primaryImageOfPage\":{\"@id\":\"https:\\\/\\\/i10x.ai\\\/news\\\/ride-rl-induced-direction-extrapolation-distillation#primaryimage\"},\"image\":{\"@id\":\"https:\\\/\\\/i10x.ai\\\/news\\\/ride-rl-induced-direction-extrapolation-distillation#primaryimage\"},\"thumbnailUrl\":\"https:\\\/\\\/i10x.ai\\\/news\\\/wp-content\\\/uploads\\\/2026\\\/10\\\/ride-rl-induced-direction-extrapolation-distillation-title.webp\",\"datePublished\":\"2026-10-19T16:49:19+00:00\",\"dateModified\":\"2026-10-19T17:20:09+00:00\",\"description\":\"RIDE distills RL teachers via representation residuals. +9.1 vs ExOPD on average; more stable than output-space extrapolation.\",\"breadcrumb\":{\"@id\":\"https:\\\/\\\/i10x.ai\\\/news\\\/ride-rl-induced-direction-extrapolation-distillation#breadcrumb\"},\"inLanguage\":\"en-US\",\"potentialAction\":[{\"@type\":\"ReadAction\",\"target\":[\"https:\\\/\\\/i10x.ai\\\/news\\\/ride-rl-induced-direction-extrapolation-distillation\"]}]},{\"@type\":\"ImageObject\",\"inLanguage\":\"en-US\",\"@id\":\"https:\\\/\\\/i10x.ai\\\/news\\\/ride-rl-induced-direction-extrapolation-distillation#primaryimage\",\"url\":\"https:\\\/\\\/i10x.ai\\\/news\\\/wp-content\\\/uploads\\\/2026\\\/10\\\/ride-rl-induced-direction-extrapolation-distillation-title.webp\",\"contentUrl\":\"https:\\\/\\\/i10x.ai\\\/news\\\/wp-content\\\/uploads\\\/2026\\\/10\\\/ride-rl-induced-direction-extrapolation-distillation-title.webp\",\"width\":1344,\"height\":768,\"caption\":\"Illustration: i10X\"},{\"@type\":\"BreadcrumbList\",\"@id\":\"https:\\\/\\\/i10x.ai\\\/news\\\/ride-rl-induced-direction-extrapolation-distillation#breadcrumb\",\"itemListElement\":[{\"@type\":\"ListItem\",\"position\":1,\"name\":\"Home\",\"item\":\"https:\\\/\\\/news.i10x.ai\\\/\"},{\"@type\":\"ListItem\",\"position\":2,\"name\":\"Distill the RL Direction in Hidden Space, Not Noisy Output Rewards\"}]},{\"@type\":\"WebSite\",\"@id\":\"https:\\\/\\\/news.i10x.ai\\\/#website\",\"url\":\"https:\\\/\\\/news.i10x.ai\\\/\",\"name\":\"i10X News\",\"description\":\"Stay updated with the latest news and announcements from i10X.\",\"publisher\":{\"@id\":\"https:\\\/\\\/news.i10x.ai\\\/#organization\"},\"potentialAction\":[{\"@type\":\"SearchAction\",\"target\":{\"@type\":\"EntryPoint\",\"urlTemplate\":\"https:\\\/\\\/news.i10x.ai\\\/?s={search_term_string}\"},\"query-input\":{\"@type\":\"PropertyValueSpecification\",\"valueRequired\":true,\"valueName\":\"search_term_string\"}}],\"inLanguage\":\"en-US\"},{\"@type\":\"Organization\",\"@id\":\"https:\\\/\\\/news.i10x.ai\\\/#organization\",\"name\":\"i10X\",\"url\":\"https:\\\/\\\/news.i10x.ai\\\/\",\"logo\":{\"@type\":\"ImageObject\",\"inLanguage\":\"en-US\",\"@id\":\"https:\\\/\\\/news.i10x.ai\\\/#\\\/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:\\\/\\\/news.i10x.ai\\\/#\\\/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:\\\/\\\/news.i10x.ai\\\/#\\\/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":"RIDE: Extrapolating RL Residuals for Better Distillation - i10X News","description":"RIDE distills RL teachers via representation residuals. +9.1 vs ExOPD on average; more stable than output-space extrapolation.","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\/ride-rl-induced-direction-extrapolation-distillation","og_locale":"en_US","og_type":"article","og_title":"RIDE: Extrapolating RL Residuals for Better Distillation - i10X News","og_description":"RIDE distills RL teachers via representation residuals. +9.1 vs ExOPD on average; more stable than output-space extrapolation.","og_url":"https:\/\/i10x.ai\/news\/ride-rl-induced-direction-extrapolation-distillation","og_site_name":"i10X News","article_published_time":"2026-10-19T16:49:19+00:00","article_modified_time":"2026-10-19T17:20:09+00:00","og_image":[{"width":1344,"height":768,"url":"https:\/\/i10x.ai\/news\/wp-content\/uploads\/2026\/10\/ride-rl-induced-direction-extrapolation-distillation-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":"2 minutes"},"schema":{"@context":"https:\/\/schema.org","@graph":[{"@type":"NewsArticle","@id":"https:\/\/i10x.ai\/news\/ride-rl-induced-direction-extrapolation-distillation#article","isPartOf":{"@id":"https:\/\/i10x.ai\/news\/ride-rl-induced-direction-extrapolation-distillation"},"author":{"name":"Christopher Ort","@id":"https:\/\/news.i10x.ai\/#\/schema\/person\/d410bb7ee72febefca2c9c369939219a"},"headline":"Distill the RL Direction in Hidden Space, Not Noisy Output Rewards","datePublished":"2026-10-19T16:49:19+00:00","dateModified":"2026-10-19T17:20:09+00:00","mainEntityOfPage":{"@id":"https:\/\/i10x.ai\/news\/ride-rl-induced-direction-extrapolation-distillation"},"wordCount":486,"commentCount":0,"publisher":{"@id":"https:\/\/news.i10x.ai\/#organization"},"image":{"@id":"https:\/\/i10x.ai\/news\/ride-rl-induced-direction-extrapolation-distillation#primaryimage"},"thumbnailUrl":"https:\/\/i10x.ai\/news\/wp-content\/uploads\/2026\/10\/ride-rl-induced-direction-extrapolation-distillation-title.webp","articleSection":["Reasoning","Studies"],"inLanguage":"en-US","potentialAction":[{"@type":"CommentAction","name":"Comment","target":["https:\/\/i10x.ai\/news\/ride-rl-induced-direction-extrapolation-distillation#respond"]}]},{"@type":"WebPage","@id":"https:\/\/i10x.ai\/news\/ride-rl-induced-direction-extrapolation-distillation","url":"https:\/\/i10x.ai\/news\/ride-rl-induced-direction-extrapolation-distillation","name":"RIDE: Extrapolating RL Residuals for Better Distillation - i10X News","isPartOf":{"@id":"https:\/\/news.i10x.ai\/#website"},"primaryImageOfPage":{"@id":"https:\/\/i10x.ai\/news\/ride-rl-induced-direction-extrapolation-distillation#primaryimage"},"image":{"@id":"https:\/\/i10x.ai\/news\/ride-rl-induced-direction-extrapolation-distillation#primaryimage"},"thumbnailUrl":"https:\/\/i10x.ai\/news\/wp-content\/uploads\/2026\/10\/ride-rl-induced-direction-extrapolation-distillation-title.webp","datePublished":"2026-10-19T16:49:19+00:00","dateModified":"2026-10-19T17:20:09+00:00","description":"RIDE distills RL teachers via representation residuals. +9.1 vs ExOPD on average; more stable than output-space extrapolation.","breadcrumb":{"@id":"https:\/\/i10x.ai\/news\/ride-rl-induced-direction-extrapolation-distillation#breadcrumb"},"inLanguage":"en-US","potentialAction":[{"@type":"ReadAction","target":["https:\/\/i10x.ai\/news\/ride-rl-induced-direction-extrapolation-distillation"]}]},{"@type":"ImageObject","inLanguage":"en-US","@id":"https:\/\/i10x.ai\/news\/ride-rl-induced-direction-extrapolation-distillation#primaryimage","url":"https:\/\/i10x.ai\/news\/wp-content\/uploads\/2026\/10\/ride-rl-induced-direction-extrapolation-distillation-title.webp","contentUrl":"https:\/\/i10x.ai\/news\/wp-content\/uploads\/2026\/10\/ride-rl-induced-direction-extrapolation-distillation-title.webp","width":1344,"height":768,"caption":"Illustration: i10X"},{"@type":"BreadcrumbList","@id":"https:\/\/i10x.ai\/news\/ride-rl-induced-direction-extrapolation-distillation#breadcrumb","itemListElement":[{"@type":"ListItem","position":1,"name":"Home","item":"https:\/\/news.i10x.ai\/"},{"@type":"ListItem","position":2,"name":"Distill the RL Direction in Hidden Space, Not Noisy Output Rewards"}]},{"@type":"WebSite","@id":"https:\/\/news.i10x.ai\/#website","url":"https:\/\/news.i10x.ai\/","name":"i10X News","description":"Stay updated with the latest news and announcements from i10X.","publisher":{"@id":"https:\/\/news.i10x.ai\/#organization"},"potentialAction":[{"@type":"SearchAction","target":{"@type":"EntryPoint","urlTemplate":"https:\/\/news.i10x.ai\/?s={search_term_string}"},"query-input":{"@type":"PropertyValueSpecification","valueRequired":true,"valueName":"search_term_string"}}],"inLanguage":"en-US"},{"@type":"Organization","@id":"https:\/\/news.i10x.ai\/#organization","name":"i10X","url":"https:\/\/news.i10x.ai\/","logo":{"@type":"ImageObject","inLanguage":"en-US","@id":"https:\/\/news.i10x.ai\/#\/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:\/\/news.i10x.ai\/#\/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:\/\/news.i10x.ai\/#\/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\/6138","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=6138"}],"version-history":[{"count":2,"href":"https:\/\/i10x.ai\/news\/wp-json\/wp\/v2\/posts\/6138\/revisions"}],"predecessor-version":[{"id":6289,"href":"https:\/\/i10x.ai\/news\/wp-json\/wp\/v2\/posts\/6138\/revisions\/6289"}],"wp:featuredmedia":[{"embeddable":true,"href":"https:\/\/i10x.ai\/news\/wp-json\/wp\/v2\/media\/6288"}],"wp:attachment":[{"href":"https:\/\/i10x.ai\/news\/wp-json\/wp\/v2\/media?parent=6138"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/i10x.ai\/news\/wp-json\/wp\/v2\/categories?post=6138"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/i10x.ai\/news\/wp-json\/wp\/v2\/tags?post=6138"},{"taxonomy":"company","embeddable":true,"href":"https:\/\/i10x.ai\/news\/wp-json\/wp\/v2\/companies?post=6138"},{"taxonomy":"product","embeddable":true,"href":"https:\/\/i10x.ai\/news\/wp-json\/wp\/v2\/products?post=6138"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}