{"id":6162,"date":"2026-10-19T16:49:37","date_gmt":"2026-10-19T16:49:37","guid":{"rendered":"https:\/\/i10xnews.kinsta.cloud\/?p=6162"},"modified":"2026-10-19T17:23:33","modified_gmt":"2026-10-19T17:23:33","slug":"realvr-grounding-latent-visual-reasoning","status":"publish","type":"post","link":"https:\/\/i10x.ai\/news\/realvr-grounding-latent-visual-reasoning","title":{"rendered":"Latent Visual Reasoning Barely Notices When the Image Changes \u2014 ReaLVR Fixes That"},"content":{"rendered":"\n<h2 class=\"wp-block-heading\">\u26a1 Quick Take<\/h2>\n\n<ul class=\"wp-block-list\"><li><strong>The finding:<\/strong> When image perturbations change the correct answer in 81.45\u201386.33% of pairs, standard LVR models flip predictions only 5.66\u201313.09% of the time \u2014 a 68\u201381 point evidence-credit gap.<\/li><li><strong>The fix:<\/strong> ReaLVR supervises free-running latent trajectories with visual-evidence contrasts under GRPO.<\/li><li><strong>The gain:<\/strong> Highest five-task averages among evaluated LVR methods: 63.7% on Qwen2.5-VL-7B, 61.2% on Qwen3-VL-8B, 65.2% on Qwen3-VL-30B.<\/li><\/ul>\n\n\n<p class=\"wp-block-paragraph\">Latent visual reasoning lets MLLMs think in continuous tokens instead of words \u2014 but those tokens are hard to supervise. University of Alabama at Birmingham and Amazon AGI researchers measure whether latent tokens actually track visual evidence.<\/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\/realvr-grounding-latent-visual-reasoning-gap.webp\" alt=\"Dumbbell ranges: ground-truth answer changes 81.45 to 86.33 percent while LVR prediction changes only 5.66 to 13.09 percent.\" class=\"wp-image-6161\" srcset=\"https:\/\/i10x.ai\/news\/wp-content\/uploads\/2026\/10\/realvr-grounding-latent-visual-reasoning-gap.webp 1344w, https:\/\/i10x.ai\/news\/wp-content\/uploads\/2026\/10\/realvr-grounding-latent-visual-reasoning-gap-300x171.webp 300w, https:\/\/i10x.ai\/news\/wp-content\/uploads\/2026\/10\/realvr-grounding-latent-visual-reasoning-gap-1024x585.webp 1024w, https:\/\/i10x.ai\/news\/wp-content\/uploads\/2026\/10\/realvr-grounding-latent-visual-reasoning-gap-768x439.webp 768w\" sizes=\"auto, (max-width: 1344px) 100vw, 1344px\" \/><figcaption class=\"wp-element-caption\">i10X chart based on data from Xiao et al. (2026), arXiv:2609.34563, Table G.5 perturbation study (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\">Final-answer rewards alone leave a credit gap. ReaLVR contrasts trajectories with evidence-aware signals. Linear probes still recover BLINK task labels at 99.9% from mean latents \u2014 states are informative but poorly grounded.<\/p>\n\n\n<ul class=\"wp-block-list\"><li><strong>Beats LVR-RL:<\/strong> +3.1 and +2.6 points five-task mean on reported backbones.<\/li><li><strong>Cross-family consistency:<\/strong> Gains hold across three model families.<\/li><\/ul>\n\n\n<h2 class=\"wp-block-heading\">\u2705 Why it matters: silent mis-reasoning<\/h2>\n\n\n<p class=\"wp-block-paragraph\">If latent tokens ignore the pixels that decide the answer, multimodal CoT is theater. ReaLVR makes evidence part of the training signal.<\/p>\n\n\n\n<h2 class=\"wp-block-heading\">\ud83e\uddea Method in brief<\/h2>\n\n\n<p class=\"wp-block-paragraph\">Perturbation analysis; GRPO with evidence contrasts; multi-backbone evals.<\/p>\n\n\n\n<h2 class=\"wp-block-heading\">\u2753 What remains open<\/h2>\n\n<ul class=\"wp-block-list\"><li><strong>Latent opacity remains:<\/strong> Even with better grounding, tokens are not human-readable.<\/li><li><strong>Task suite:<\/strong> Five-task averages; broader coverage TBD.<\/li><\/ul>\n\n\n<h2 class=\"wp-block-heading\">\ud83e\udded Context<\/h2>\n\n\n<p class=\"wp-block-paragraph\">Related multimodal reasoning in <a href=\"\/news\/category\/studies\/multimodal\">Multimodal<\/a> and <a href=\"\/news\/category\/studies\/reasoning\">Reasoning<\/a>.<\/p>\n\n\n\n<h2 class=\"wp-block-heading\">\ud83d\udee0\ufe0f What it means for MLLM trainers<\/h2>\n\n<ul class=\"wp-block-list\"><li><strong>Test evidence sensitivity:<\/strong> Perturb images that flip labels.<\/li><li><strong>Supervise latents:<\/strong> Answer reward is not enough.<\/li><li><strong>Probe representations:<\/strong> High probe accuracy \u2260 grounded use.<\/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;Rethinking Latent Visual Reasoning: Grounding Latent Reasoning in Visual Evidence&#8221; (arXiv:2609.34563v1, submitted September 28, 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 Table G.5 perturbation study.<\/p>\n\n\n\n<h2 class=\"wp-block-heading\">\ud83d\udd2d i10x Perspective<\/h2>\n\n\n<p class=\"wp-block-paragraph\">An 80-point gap between what should change and what does is devastating for latent reasoning hype. ReaLVR&#8217;s contribution is insisting that continuous thoughts still need visual receipts.<\/p>\n\n","protected":false},"excerpt":{"rendered":"<p>Amazon\/UAB study: LVR models change predictions only 5.7\u201313.1% when image edits flip the answer 81\u201386% of the time. ReaLVR hits 63.7% five-task average on Qwen2.5-VL-7B.<\/p>\n","protected":false},"author":2,"featured_media":6312,"comment_status":"open","ping_status":"open","sticky":false,"template":"","format":"standard","meta":{"paper_title":"Rethinking Latent Visual Reasoning: Grounding Latent Reasoning in Visual Evidence","paper_url":"https:\/\/arxiv.org\/abs\/2609.34563","pdf_url":"https:\/\/arxiv.org\/pdf\/2609.34563v1","doi":"","arxiv_id":"2609.34563","authors":"Xi Xiao, Tianchen Zhao, Youngeun Kim, Zhuowei Li, Linghan Xu, Jiaye Wu, Zheng Zhang, Xiang Xu, Xuanbai Chen, Farhan Tejani, Jakub Zablocki, Julia Xu, Yifan Xing","institutions":"University of Alabama at Birmingham; Amazon AGI","paper_date":"2026-09-28","venue":"arXiv (cs.CV), version 1","peer_reviewed":false,"code_url":"","data_url":"","funding_coi":"Amazon AGI co-authors.","figure_license":"Chart drawn by i10X from Table G.5 perturbation study (arXiv non-exclusive license)","paper_license":"arXiv non-exclusive license","replication":"no","study_models":"Qwen2.5-VL-7B, Qwen3-VL-8B\/30B and other LVR baselines","study_sample":"Five-task latent-reasoning suite plus perturbation analyses","footnotes":"","_yoast_wpseo_primary_category":""},"categories":[58,52],"tags":[],"companies":[],"products":[],"class_list":["post-6162","post","type-post","status-publish","format-standard","has-post-thumbnail","hentry","category-multimodal","category-studies"],"yoast_head":"<!-- This site is optimized with the Yoast SEO plugin v28.7 - https:\/\/yoast.com\/product\/yoast-seo-wordpress\/ -->\n<title>ReaLVR: Grounding Latent Reasoning in Visual Evidence - i10X News<\/title>\n<meta name=\"description\" content=\"Latent visual reasoning ignores image edits that change answers. ReaLVR adds evidence supervision; 63.7% five-task avg on Qwen2.5-VL-7B.\" \/>\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\/realvr-grounding-latent-visual-reasoning\" \/>\n<meta property=\"og:locale\" content=\"en_US\" \/>\n<meta property=\"og:type\" content=\"article\" \/>\n<meta property=\"og:title\" content=\"ReaLVR: Grounding Latent Reasoning in Visual Evidence - i10X News\" \/>\n<meta property=\"og:description\" content=\"Latent visual reasoning ignores image edits that change answers. ReaLVR adds evidence supervision; 63.7% five-task avg on Qwen2.5-VL-7B.\" \/>\n<meta property=\"og:url\" content=\"https:\/\/i10x.ai\/news\/realvr-grounding-latent-visual-reasoning\" \/>\n<meta property=\"og:site_name\" content=\"i10X News\" \/>\n<meta property=\"article:published_time\" content=\"2026-10-19T16:49:37+00:00\" \/>\n<meta property=\"article:modified_time\" content=\"2026-10-19T17:23:33+00:00\" \/>\n<meta property=\"og:image\" content=\"https:\/\/i10x.ai\/news\/wp-content\/uploads\/2026\/10\/realvr-grounding-latent-visual-reasoning-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\\\/realvr-grounding-latent-visual-reasoning#article\",\"isPartOf\":{\"@id\":\"https:\\\/\\\/i10x.ai\\\/news\\\/realvr-grounding-latent-visual-reasoning\"},\"author\":{\"name\":\"Christopher Ort\",\"@id\":\"https:\\\/\\\/news.i10x.ai\\\/#\\\/schema\\\/person\\\/d410bb7ee72febefca2c9c369939219a\"},\"headline\":\"Latent Visual Reasoning Barely Notices When the Image Changes \u2014 ReaLVR Fixes That\",\"datePublished\":\"2026-10-19T16:49:37+00:00\",\"dateModified\":\"2026-10-19T17:23:33+00:00\",\"mainEntityOfPage\":{\"@id\":\"https:\\\/\\\/i10x.ai\\\/news\\\/realvr-grounding-latent-visual-reasoning\"},\"wordCount\":382,\"commentCount\":0,\"publisher\":{\"@id\":\"https:\\\/\\\/news.i10x.ai\\\/#organization\"},\"image\":{\"@id\":\"https:\\\/\\\/i10x.ai\\\/news\\\/realvr-grounding-latent-visual-reasoning#primaryimage\"},\"thumbnailUrl\":\"https:\\\/\\\/i10x.ai\\\/news\\\/wp-content\\\/uploads\\\/2026\\\/10\\\/realvr-grounding-latent-visual-reasoning-title.webp\",\"articleSection\":[\"Multimodal\",\"Studies\"],\"inLanguage\":\"en-US\",\"potentialAction\":[{\"@type\":\"CommentAction\",\"name\":\"Comment\",\"target\":[\"https:\\\/\\\/i10x.ai\\\/news\\\/realvr-grounding-latent-visual-reasoning#respond\"]}]},{\"@type\":\"WebPage\",\"@id\":\"https:\\\/\\\/i10x.ai\\\/news\\\/realvr-grounding-latent-visual-reasoning\",\"url\":\"https:\\\/\\\/i10x.ai\\\/news\\\/realvr-grounding-latent-visual-reasoning\",\"name\":\"ReaLVR: Grounding Latent Reasoning in Visual Evidence - i10X News\",\"isPartOf\":{\"@id\":\"https:\\\/\\\/news.i10x.ai\\\/#website\"},\"primaryImageOfPage\":{\"@id\":\"https:\\\/\\\/i10x.ai\\\/news\\\/realvr-grounding-latent-visual-reasoning#primaryimage\"},\"image\":{\"@id\":\"https:\\\/\\\/i10x.ai\\\/news\\\/realvr-grounding-latent-visual-reasoning#primaryimage\"},\"thumbnailUrl\":\"https:\\\/\\\/i10x.ai\\\/news\\\/wp-content\\\/uploads\\\/2026\\\/10\\\/realvr-grounding-latent-visual-reasoning-title.webp\",\"datePublished\":\"2026-10-19T16:49:37+00:00\",\"dateModified\":\"2026-10-19T17:23:33+00:00\",\"description\":\"Latent visual reasoning ignores image edits that change answers. ReaLVR adds evidence supervision; 63.7% five-task avg on Qwen2.5-VL-7B.\",\"breadcrumb\":{\"@id\":\"https:\\\/\\\/i10x.ai\\\/news\\\/realvr-grounding-latent-visual-reasoning#breadcrumb\"},\"inLanguage\":\"en-US\",\"potentialAction\":[{\"@type\":\"ReadAction\",\"target\":[\"https:\\\/\\\/i10x.ai\\\/news\\\/realvr-grounding-latent-visual-reasoning\"]}]},{\"@type\":\"ImageObject\",\"inLanguage\":\"en-US\",\"@id\":\"https:\\\/\\\/i10x.ai\\\/news\\\/realvr-grounding-latent-visual-reasoning#primaryimage\",\"url\":\"https:\\\/\\\/i10x.ai\\\/news\\\/wp-content\\\/uploads\\\/2026\\\/10\\\/realvr-grounding-latent-visual-reasoning-title.webp\",\"contentUrl\":\"https:\\\/\\\/i10x.ai\\\/news\\\/wp-content\\\/uploads\\\/2026\\\/10\\\/realvr-grounding-latent-visual-reasoning-title.webp\",\"width\":1344,\"height\":768,\"caption\":\"Illustration: i10X\"},{\"@type\":\"BreadcrumbList\",\"@id\":\"https:\\\/\\\/i10x.ai\\\/news\\\/realvr-grounding-latent-visual-reasoning#breadcrumb\",\"itemListElement\":[{\"@type\":\"ListItem\",\"position\":1,\"name\":\"Home\",\"item\":\"https:\\\/\\\/news.i10x.ai\\\/\"},{\"@type\":\"ListItem\",\"position\":2,\"name\":\"Latent Visual Reasoning Barely Notices When the Image Changes \u2014 ReaLVR Fixes That\"}]},{\"@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":"ReaLVR: Grounding Latent Reasoning in Visual Evidence - i10X News","description":"Latent visual reasoning ignores image edits that change answers. ReaLVR adds evidence supervision; 63.7% five-task avg on Qwen2.5-VL-7B.","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\/realvr-grounding-latent-visual-reasoning","og_locale":"en_US","og_type":"article","og_title":"ReaLVR: Grounding Latent Reasoning in Visual Evidence - i10X News","og_description":"Latent visual reasoning ignores image edits that change answers. ReaLVR adds evidence supervision; 63.7% five-task avg on Qwen2.5-VL-7B.","og_url":"https:\/\/i10x.ai\/news\/realvr-grounding-latent-visual-reasoning","og_site_name":"i10X News","article_published_time":"2026-10-19T16:49:37+00:00","article_modified_time":"2026-10-19T17:23:33+00:00","og_image":[{"width":1344,"height":768,"url":"https:\/\/i10x.ai\/news\/wp-content\/uploads\/2026\/10\/realvr-grounding-latent-visual-reasoning-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\/realvr-grounding-latent-visual-reasoning#article","isPartOf":{"@id":"https:\/\/i10x.ai\/news\/realvr-grounding-latent-visual-reasoning"},"author":{"name":"Christopher Ort","@id":"https:\/\/news.i10x.ai\/#\/schema\/person\/d410bb7ee72febefca2c9c369939219a"},"headline":"Latent Visual Reasoning Barely Notices When the Image Changes \u2014 ReaLVR Fixes That","datePublished":"2026-10-19T16:49:37+00:00","dateModified":"2026-10-19T17:23:33+00:00","mainEntityOfPage":{"@id":"https:\/\/i10x.ai\/news\/realvr-grounding-latent-visual-reasoning"},"wordCount":382,"commentCount":0,"publisher":{"@id":"https:\/\/news.i10x.ai\/#organization"},"image":{"@id":"https:\/\/i10x.ai\/news\/realvr-grounding-latent-visual-reasoning#primaryimage"},"thumbnailUrl":"https:\/\/i10x.ai\/news\/wp-content\/uploads\/2026\/10\/realvr-grounding-latent-visual-reasoning-title.webp","articleSection":["Multimodal","Studies"],"inLanguage":"en-US","potentialAction":[{"@type":"CommentAction","name":"Comment","target":["https:\/\/i10x.ai\/news\/realvr-grounding-latent-visual-reasoning#respond"]}]},{"@type":"WebPage","@id":"https:\/\/i10x.ai\/news\/realvr-grounding-latent-visual-reasoning","url":"https:\/\/i10x.ai\/news\/realvr-grounding-latent-visual-reasoning","name":"ReaLVR: Grounding Latent Reasoning in Visual Evidence - i10X News","isPartOf":{"@id":"https:\/\/news.i10x.ai\/#website"},"primaryImageOfPage":{"@id":"https:\/\/i10x.ai\/news\/realvr-grounding-latent-visual-reasoning#primaryimage"},"image":{"@id":"https:\/\/i10x.ai\/news\/realvr-grounding-latent-visual-reasoning#primaryimage"},"thumbnailUrl":"https:\/\/i10x.ai\/news\/wp-content\/uploads\/2026\/10\/realvr-grounding-latent-visual-reasoning-title.webp","datePublished":"2026-10-19T16:49:37+00:00","dateModified":"2026-10-19T17:23:33+00:00","description":"Latent visual reasoning ignores image edits that change answers. ReaLVR adds evidence supervision; 63.7% five-task avg on Qwen2.5-VL-7B.","breadcrumb":{"@id":"https:\/\/i10x.ai\/news\/realvr-grounding-latent-visual-reasoning#breadcrumb"},"inLanguage":"en-US","potentialAction":[{"@type":"ReadAction","target":["https:\/\/i10x.ai\/news\/realvr-grounding-latent-visual-reasoning"]}]},{"@type":"ImageObject","inLanguage":"en-US","@id":"https:\/\/i10x.ai\/news\/realvr-grounding-latent-visual-reasoning#primaryimage","url":"https:\/\/i10x.ai\/news\/wp-content\/uploads\/2026\/10\/realvr-grounding-latent-visual-reasoning-title.webp","contentUrl":"https:\/\/i10x.ai\/news\/wp-content\/uploads\/2026\/10\/realvr-grounding-latent-visual-reasoning-title.webp","width":1344,"height":768,"caption":"Illustration: i10X"},{"@type":"BreadcrumbList","@id":"https:\/\/i10x.ai\/news\/realvr-grounding-latent-visual-reasoning#breadcrumb","itemListElement":[{"@type":"ListItem","position":1,"name":"Home","item":"https:\/\/news.i10x.ai\/"},{"@type":"ListItem","position":2,"name":"Latent Visual Reasoning Barely Notices When the Image Changes \u2014 ReaLVR Fixes That"}]},{"@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\/6162","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=6162"}],"version-history":[{"count":2,"href":"https:\/\/i10x.ai\/news\/wp-json\/wp\/v2\/posts\/6162\/revisions"}],"predecessor-version":[{"id":6313,"href":"https:\/\/i10x.ai\/news\/wp-json\/wp\/v2\/posts\/6162\/revisions\/6313"}],"wp:featuredmedia":[{"embeddable":true,"href":"https:\/\/i10x.ai\/news\/wp-json\/wp\/v2\/media\/6312"}],"wp:attachment":[{"href":"https:\/\/i10x.ai\/news\/wp-json\/wp\/v2\/media?parent=6162"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/i10x.ai\/news\/wp-json\/wp\/v2\/categories?post=6162"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/i10x.ai\/news\/wp-json\/wp\/v2\/tags?post=6162"},{"taxonomy":"company","embeddable":true,"href":"https:\/\/i10x.ai\/news\/wp-json\/wp\/v2\/companies?post=6162"},{"taxonomy":"product","embeddable":true,"href":"https:\/\/i10x.ai\/news\/wp-json\/wp\/v2\/products?post=6162"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}