{"id":49375,"date":"2026-11-18T10:39:00","date_gmt":"2026-11-18T10:39:00","guid":{"rendered":"https:\/\/en.prnasia.com\/story\/archive\/5101074_CN01074_0"},"modified":"2026-11-18T10:39:00","modified_gmt":"2026-11-18T10:39:00","slug":"einsia-ai-launches-visual-cipher-bench-to-evaluate-long-horizon-visual-reasoning-in-ai-agents","status":"publish","type":"post","link":"https:\/\/transmediavictoria.net.au\/?p=49375","title":{"rendered":"Einsia AI Launches Visual Cipher Bench to Evaluate Long-Horizon Visual Reasoning in AI Agents"},"content":{"rendered":"<div id=\"dvContent\" readability=\"159.78431697613\">\n<p class=\"prntal\"><b>Top-performing model achieves just 26.7% accuracy, revealing a significant gap in autonomous rule discovery<\/b><\/p>\n<p id=\"temp_ReleaseStart\"><span class=\"legendSpanClass\">SINGAPORE<\/span>, <span class=\"legendSpanClass\">Nov. 18, 2026<\/span> \/PRNewswire\/ &#8212; As AI agents evolve from answering questions to completing complex tasks, a fundamental challenge remains: Can they solve problems when instructions are absent and rules must be discovered entirely from visual clues?<\/p>\n<p class=\"prntal\"><b>Einsia AI<\/b>&#8216;s Navers Lab today announced <b>Visual Cipher Bench<\/b>, a new benchmark evaluating long-horizon visual reasoning in AI agents, focusing on rule discovery, logical planning, and spatial operations.<\/p>\n<p class=\"prntal\">Across <b>eight frontier models <\/b>and <b>29 reasoning configurations<\/b>, the benchmark recorded <b>3,480 attempts<\/b>. No configuration achieved accuracy above 30%. The top performer, <b>GPT-6 Astra (max), scored 26.7%<\/b>. DeepSeek V4.1 Flash, Qwen 3.8 Max, and Kimi K3.1 Pro failed to solve any puzzles.<\/p>\n<p class=\"prntal\">The findings expose a critical gap between executing known instructions and independently discovering unknown rules. While spatial reasoning proved relatively manageable, accuracy on rule induction was just 0.6%.<\/p>\n<div class=\"PRN_ImbeddedAssetReference\" id=\"DivAssetPlaceHolder5698\" readability=\"11\">\n<p><a href=\"https:\/\/mmx.prnasia.com\/media\/MS2038583\/20261118012432EST_image_3.jpg?id=OA3087316&amp;p=medium600\" target=\"_blank\"><img decoding=\"async\" src=\"https:\/\/mmx.prnasia.com\/media\/MS2038583\/20261118012432EST_image_3.jpg?id=OA3087316&amp;p=medium600\" title=\"Figure 1: Overview of Visual Cipher Bench. Left: 40 original puzzles (67 images) in four categories. Middle: each model's best configuration, averaged over three runs. Right: the evaluation pipeline, from puzzle images to a single scored answer.\" alt=\"Figure 1: Overview of Visual Cipher Bench. Left: 40 original puzzles (67 images) in four categories. Middle: each model's best configuration, averaged over three runs. Right: the evaluation pipeline, from puzzle images to a single scored answer.\"><\/a><br \/><span>Figure 1: Overview of Visual Cipher Bench. Left: 40 original puzzles (67 images) in four categories. Middle: each model&#8217;s best configuration, averaged over three runs. Right: the evaluation pipeline, from puzzle images to a single scored answer.<\/span><\/p>\n<\/p><\/div>\n<p class=\"prntal\"><b>Testing AI Beyond Conventional Visual Question Answering<\/b><\/p>\n<p class=\"prntal\">Unlike conventional visual benchmarks, which typically provide explicit questions and expected answers, Visual Cipher Bench tests scenarios where <b>instructions are hidden in images, with no explicit rules or immediate correctness feedback<\/b>.<\/p>\n<p class=\"prntal\">The benchmark comprises 40 original visual puzzles represented by 67 black-and-white illustrations. Agents receive one instruction: examine the images and write the final answer to a file. They must independently identify examples, understand the target problem, infer the required output, and solve each task.<\/p>\n<p class=\"prntal\">The puzzles cover four categories:<\/p>\n<ul type=\"disc\">\n<li><b>Spatial Operations:<\/b> Folding, rotation, projection, and geometric transformations.<\/li>\n<li><b>Logical Planning:<\/b> Extracting visual constraints and finding valid solutions.<\/li>\n<li><b>Rule Induction:<\/b> Discovering hidden patterns from limited examples.<\/li>\n<li><b>Knowledge Decoding:<\/b> Interpreting Morse code, Braille, and word puzzles.<\/li>\n<\/ul>\n<p class=\"prntal\"><b>Frontier Models Struggle with Autonomous Rule Discovery<\/b><\/p>\n<p class=\"prntal\">Performance varied significantly across the eight evaluated models:<\/p>\n<div class=\"PRN_ImbeddedAssetReference\" id=\"DivAssetPlaceHolder6399\" readability=\"9\">\n<p><a href=\"https:\/\/mmx.prnasia.com\/media\/MS2038582\/20261118012432EST_image_2.jpg?id=OA3087315&amp;p=medium600\" target=\"_blank\"><img decoding=\"async\" src=\"https:\/\/mmx.prnasia.com\/media\/MS2038582\/20261118012432EST_image_2.jpg?id=OA3087315&amp;p=medium600\" title=\"Figure 2: Accuracy (%) of the eight evaluated models. Each bar is one reasoning-effort setting, from lowest to highest; the number shown is the model's best configuration.\" alt=\"Figure 2: Accuracy (%) of the eight evaluated models. Each bar is one reasoning-effort setting, from lowest to highest; the number shown is the model's best configuration.\"><\/a><br \/><span>Figure 2: Accuracy (%) of the eight evaluated models. Each bar is one reasoning-effort setting, from lowest to highest; the number shown is the model&#8217;s best configuration.<\/span><\/p>\n<\/p><\/div>\n<ul type=\"disc\">\n<li><b>GPT-6 Astra (max):<\/b> 26.7%, averaging 10.7 correct answers per 40-puzzle run.<\/li>\n<li><b>GPT-6.1 Sol (max) and Claude Opus 5.5 (xhigh):<\/b> Both scored 21.7%.<\/li>\n<li><b>GPT-6 Luna and Claude Sonnet 5.5:<\/b> Both scored 0.8%, solving one puzzle each across 120 attempts.<\/li>\n<li><b>DeepSeek V4.1 Flash, Qwen 3.8 Max, and Kimi K3.1 Pro:<\/b> Zero correct answers.<\/li>\n<\/ul>\n<p class=\"prntal\">Across all 3,480 attempts, accuracy by category was:<\/p>\n<ul type=\"disc\">\n<li>Spatial Operations: 22.8%<\/li>\n<li>Knowledge Decoding: 3.3%<\/li>\n<li>Logical Planning: 2.4%<\/li>\n<li><b>Rule Induction: 0.6%<\/b><\/li>\n<\/ul>\n<p class=\"prntal\">Models solved only five of 870 rule-induction attempts. Even GPT-6 Astra (max) achieved 80% on spatial operations, while no configuration exceeded 13.3% in the other three categories. Researchers also found that some correct answers resulted from thematic guesses rather than systematic rule discovery.<\/p>\n<div class=\"PRN_ImbeddedAssetReference\" id=\"DivAssetPlaceHolder3698\" readability=\"9\">\n<p><a href=\"https:\/\/mmx.prnasia.com\/media\/MS2038584\/20261118012432EST_image_4.jpg?id=OA3087317&amp;p=medium600\" target=\"_blank\"><img decoding=\"async\" src=\"https:\/\/mmx.prnasia.com\/media\/MS2038584\/20261118012432EST_image_4.jpg?id=OA3087317&amp;p=medium600\" title=\"Figure 3: Full results for all 29 model\u2013effort configurations, with accuracy by category and overall. Shaded rows mark each model's best configuration.\" alt=\"Figure 3: Full results for all 29 model\u2013effort configurations, with accuracy by category and overall. Shaded rows mark each model's best configuration.\"><\/a><br \/><span>Figure 3: Full results for all 29 model\u2013effort configurations, with accuracy by category and overall. Shaded rows mark each model&#8217;s best configuration.<\/span><\/p>\n<\/p><\/div>\n<p class=\"prntal\"><b>Providing Rules Raises Accuracy from 12.2% to 51.1%<\/b><\/p>\n<p class=\"prntal\">To distinguish rule discovery from rule execution, researchers conducted an additional experiment on 10 puzzles using GPT-6 Astra, GPT-6.1 Sol, and GPT-6 Luna.<\/p>\n<div class=\"PRN_ImbeddedAssetReference\" id=\"DivAssetPlaceHolder2454\" readability=\"9\">\n<p><a href=\"https:\/\/mmx.prnasia.com\/media\/MS2038581\/20261118012432EST_image_1.jpg?id=OA3087314&amp;p=medium600\" target=\"_blank\"><img decoding=\"async\" src=\"https:\/\/mmx.prnasia.com\/media\/MS2038581\/20261118012432EST_image_1.jpg?id=OA3087314&amp;p=medium600\" title=\"Figure 4: Hint experiment on 10 puzzles with three models. A one-sentence mechanism hint raises overall accuracy from 12.2% to 51.1%, and rule induction from 0% to 85.2%.\" alt=\"Figure 4: Hint experiment on 10 puzzles with three models. A one-sentence mechanism hint raises overall accuracy from 12.2% to 51.1%, and rule induction from 0% to 85.2%.\"><\/a><br \/><span>Figure 4: Hint experiment on 10 puzzles with three models. A one-sentence mechanism hint raises overall accuracy from 12.2% to 51.1%, and rule induction from 0% to 85.2%.<\/span><\/p>\n<\/p><\/div>\n<p class=\"prntal\">When provided with the core rules\u2014but not intermediate reasoning steps or answers\u2014overall accuracy increased from <b>12.2% to 51.1%<\/b>. Rule-induction accuracy rose from 0% to 85.2%.<\/p>\n<p class=\"prntal\">GPT-6 Astra improved from 23.3% to 70.0%, while GPT-6.1 Sol increased from 13.3% to 66.7%. Timeout occurrences also fell from 51 to 29 across 90 attempts.<\/p>\n<p class=\"prntal\">A control experiment providing additional information about task-state representation, but not the core rules, achieved only 16.7% accuracy. Even with rule hints, logical planning accuracy remained at 22.2%.<\/p>\n<p class=\"prntal\"><b>Why Do AI Agents Fail?<\/b><\/p>\n<p class=\"prntal\">Analysis of failure trajectories revealed two recurring patterns.<\/p>\n<p class=\"prntal\"><b>Premature conclusions:<\/b> Agents often adopted rules that explained only some examples without validating them against all available evidence. Some submitted answers despite acknowledging incomplete understanding.<\/p>\n<p class=\"prntal\"><b>Inefficient search:<\/b> Agents frequently relied on extensive tool calls and exhaustive combinations instead of using visual constraints to narrow the search space, eventually timing out.<\/p>\n<p class=\"prntal\">More effective reasoning involved generating multiple hypotheses, testing them against examples, eliminating inconsistent rules, and applying the remaining valid hypothesis.<\/p>\n<p class=\"prntal\"><b>From Following Instructions to Discovering Rules<\/b><\/p>\n<p class=\"prntal\">In software development, agents typically receive explicit requirements and can rely on test results to evaluate their solutions. Real-world problems often provide neither.<\/p>\n<p class=\"prntal\">Agents must observe their environment, identify relevant signals, formulate hypotheses, and validate their understanding without explicit instructions or immediate feedback.<\/p>\n<p class=\"prntal\"><b>Visual Cipher Bench<\/b> provides a systematic evaluation of this capability gap. Its findings suggest that independently discovering, validating, and transferring hidden rules remains a significant challenge for frontier AI agents \u2014a capability increasingly important as AI moves toward more autonomous problem-solving.<\/p>\n<p class=\"prntal\"><b>Research and Resources<\/b><\/p>\n<\/p><\/div>\n<p>Source: Einsia AI Technology (SG) PTE. LTD.<\/p>\n","protected":false},"excerpt":{"rendered":"<p>Top-performing model achieves just 26.7% accuracy, revealing a significant gap in autonomous rule discovery SINGAPORE, Nov. 18, 2026 \/PRNewswire\/ &#8212; As AI agents evolve from answering questions to completing complex tasks, a fundamental challenge remains: Can they solve problems when instructions are absent and rules must be discovered entirely from visual clues? Einsia AI&#8216;s Navers [&hellip;]<\/p>\n","protected":false},"author":8,"featured_media":49376,"comment_status":"closed","ping_status":"closed","sticky":false,"template":"","format":"standard","meta":{"footnotes":""},"categories":[20],"tags":[],"class_list":["post-49375","post","type-post","status-publish","format-standard","has-post-thumbnail","hentry","category-business-news"],"yoast_head":"<!-- This site is optimized with the Yoast SEO plugin v27.9 - https:\/\/yoast.com\/product\/yoast-seo-wordpress\/ -->\n<title>Einsia AI Launches Visual Cipher Bench to Evaluate Long-Horizon Visual Reasoning in AI Agents - My Blog<\/title>\n<meta name=\"robots\" content=\"index, follow, max-snippet:-1, max-image-preview:large, max-video-preview:-1\" \/>\n<link rel=\"canonical\" href=\"https:\/\/transmediavictoria.net.au\/?p=49375\" \/>\n<meta property=\"og:locale\" content=\"en_US\" \/>\n<meta property=\"og:type\" content=\"article\" \/>\n<meta property=\"og:title\" content=\"Einsia AI Launches Visual Cipher Bench to Evaluate Long-Horizon Visual Reasoning in AI Agents - My Blog\" \/>\n<meta property=\"og:description\" content=\"Top-performing model achieves just 26.7% accuracy, revealing a significant gap in autonomous rule discovery SINGAPORE, Nov. 18, 2026 \/PRNewswire\/ &#8212; As AI agents evolve from answering questions to completing complex tasks, a fundamental challenge remains: Can they solve problems when instructions are absent and rules must be discovered entirely from visual clues? Einsia AI&#8216;s Navers [&hellip;]\" \/>\n<meta property=\"og:url\" content=\"https:\/\/transmediavictoria.net.au\/?p=49375\" \/>\n<meta property=\"og:site_name\" content=\"My Blog\" \/>\n<meta property=\"article:published_time\" content=\"2026-11-18T10:39:00+00:00\" \/>\n<meta property=\"og:image\" content=\"https:\/\/mmx.prnasia.com\/media\/MS2038583\/20261118012432EST_image_3.jpg?id=OA3087316&amp;p=medium600\" \/>\n<meta name=\"author\" content=\"admin_tony\" \/>\n<meta name=\"twitter:card\" content=\"summary_large_image\" \/>\n<meta name=\"twitter:label1\" content=\"Written by\" \/>\n\t<meta name=\"twitter:data1\" content=\"admin_tony\" \/>\n\t<meta name=\"twitter:label2\" content=\"Est. reading time\" \/>\n\t<meta name=\"twitter:data2\" content=\"4 minutes\" \/>\n<script type=\"application\/ld+json\" class=\"yoast-schema-graph\">{\"@context\":\"https:\\\/\\\/schema.org\",\"@graph\":[{\"@type\":\"Article\",\"@id\":\"https:\\\/\\\/transmediavictoria.net.au\\\/?p=49375#article\",\"isPartOf\":{\"@id\":\"https:\\\/\\\/transmediavictoria.net.au\\\/?p=49375\"},\"author\":{\"name\":\"admin_tony\",\"@id\":\"https:\\\/\\\/transmediavictoria.net.au\\\/#\\\/schema\\\/person\\\/5b51d407d652c0aa802bea269ea5d7e7\"},\"headline\":\"Einsia AI Launches Visual Cipher Bench to Evaluate Long-Horizon Visual Reasoning in AI Agents\",\"datePublished\":\"2026-11-18T10:39:00+00:00\",\"mainEntityOfPage\":{\"@id\":\"https:\\\/\\\/transmediavictoria.net.au\\\/?p=49375\"},\"wordCount\":768,\"image\":{\"@id\":\"https:\\\/\\\/transmediavictoria.net.au\\\/?p=49375#primaryimage\"},\"thumbnailUrl\":\"https:\\\/\\\/transmediavictoria.net.au\\\/wp-content\\\/uploads\\\/2026\\\/11\\\/einsia-ai-launches-visual-cipher-bench-to-evaluate-long-horizon-visual-reasoning-in-ai-agents.jpg\",\"articleSection\":[\"Business News\"],\"inLanguage\":\"en-US\"},{\"@type\":\"WebPage\",\"@id\":\"https:\\\/\\\/transmediavictoria.net.au\\\/?p=49375\",\"url\":\"https:\\\/\\\/transmediavictoria.net.au\\\/?p=49375\",\"name\":\"Einsia AI Launches Visual Cipher Bench to Evaluate Long-Horizon Visual Reasoning in AI Agents - My Blog\",\"isPartOf\":{\"@id\":\"https:\\\/\\\/transmediavictoria.net.au\\\/#website\"},\"primaryImageOfPage\":{\"@id\":\"https:\\\/\\\/transmediavictoria.net.au\\\/?p=49375#primaryimage\"},\"image\":{\"@id\":\"https:\\\/\\\/transmediavictoria.net.au\\\/?p=49375#primaryimage\"},\"thumbnailUrl\":\"https:\\\/\\\/transmediavictoria.net.au\\\/wp-content\\\/uploads\\\/2026\\\/11\\\/einsia-ai-launches-visual-cipher-bench-to-evaluate-long-horizon-visual-reasoning-in-ai-agents.jpg\",\"datePublished\":\"2026-11-18T10:39:00+00:00\",\"author\":{\"@id\":\"https:\\\/\\\/transmediavictoria.net.au\\\/#\\\/schema\\\/person\\\/5b51d407d652c0aa802bea269ea5d7e7\"},\"breadcrumb\":{\"@id\":\"https:\\\/\\\/transmediavictoria.net.au\\\/?p=49375#breadcrumb\"},\"inLanguage\":\"en-US\",\"potentialAction\":[{\"@type\":\"ReadAction\",\"target\":[\"https:\\\/\\\/transmediavictoria.net.au\\\/?p=49375\"]}]},{\"@type\":\"ImageObject\",\"inLanguage\":\"en-US\",\"@id\":\"https:\\\/\\\/transmediavictoria.net.au\\\/?p=49375#primaryimage\",\"url\":\"https:\\\/\\\/transmediavictoria.net.au\\\/wp-content\\\/uploads\\\/2026\\\/11\\\/einsia-ai-launches-visual-cipher-bench-to-evaluate-long-horizon-visual-reasoning-in-ai-agents.jpg\",\"contentUrl\":\"https:\\\/\\\/transmediavictoria.net.au\\\/wp-content\\\/uploads\\\/2026\\\/11\\\/einsia-ai-launches-visual-cipher-bench-to-evaluate-long-horizon-visual-reasoning-in-ai-agents.jpg\",\"width\":599,\"height\":268},{\"@type\":\"BreadcrumbList\",\"@id\":\"https:\\\/\\\/transmediavictoria.net.au\\\/?p=49375#breadcrumb\",\"itemListElement\":[{\"@type\":\"ListItem\",\"position\":1,\"name\":\"Home\",\"item\":\"https:\\\/\\\/transmediavictoria.net.au\\\/\"},{\"@type\":\"ListItem\",\"position\":2,\"name\":\"Einsia AI Launches Visual Cipher Bench to Evaluate Long-Horizon Visual Reasoning in AI Agents\"}]},{\"@type\":\"WebSite\",\"@id\":\"https:\\\/\\\/transmediavictoria.net.au\\\/#website\",\"url\":\"https:\\\/\\\/transmediavictoria.net.au\\\/\",\"name\":\"My Blog\",\"description\":\"My WordPress Blog\",\"potentialAction\":[{\"@type\":\"SearchAction\",\"target\":{\"@type\":\"EntryPoint\",\"urlTemplate\":\"https:\\\/\\\/transmediavictoria.net.au\\\/?s={search_term_string}\"},\"query-input\":{\"@type\":\"PropertyValueSpecification\",\"valueRequired\":true,\"valueName\":\"search_term_string\"}}],\"inLanguage\":\"en-US\"},{\"@type\":\"Person\",\"@id\":\"https:\\\/\\\/transmediavictoria.net.au\\\/#\\\/schema\\\/person\\\/5b51d407d652c0aa802bea269ea5d7e7\",\"name\":\"admin_tony\",\"image\":{\"@type\":\"ImageObject\",\"inLanguage\":\"en-US\",\"@id\":\"https:\\\/\\\/secure.gravatar.com\\\/avatar\\\/102836fe5f9cd6c64e96e90e7341dbc815b7d12726a85708326dc0db40afd906?s=96&d=mm&r=g\",\"url\":\"https:\\\/\\\/secure.gravatar.com\\\/avatar\\\/102836fe5f9cd6c64e96e90e7341dbc815b7d12726a85708326dc0db40afd906?s=96&d=mm&r=g\",\"contentUrl\":\"https:\\\/\\\/secure.gravatar.com\\\/avatar\\\/102836fe5f9cd6c64e96e90e7341dbc815b7d12726a85708326dc0db40afd906?s=96&d=mm&r=g\",\"caption\":\"admin_tony\"},\"url\":\"https:\\\/\\\/transmediavictoria.net.au\\\/?author=8\"}]}<\/script>\n<!-- \/ Yoast SEO plugin. -->","yoast_head_json":{"title":"Einsia AI Launches Visual Cipher Bench to Evaluate Long-Horizon Visual Reasoning in AI Agents - My Blog","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:\/\/transmediavictoria.net.au\/?p=49375","og_locale":"en_US","og_type":"article","og_title":"Einsia AI Launches Visual Cipher Bench to Evaluate Long-Horizon Visual Reasoning in AI Agents - My Blog","og_description":"Top-performing model achieves just 26.7% accuracy, revealing a significant gap in autonomous rule discovery SINGAPORE, Nov. 18, 2026 \/PRNewswire\/ &#8212; As AI agents evolve from answering questions to completing complex tasks, a fundamental challenge remains: Can they solve problems when instructions are absent and rules must be discovered entirely from visual clues? Einsia AI&#8216;s Navers [&hellip;]","og_url":"https:\/\/transmediavictoria.net.au\/?p=49375","og_site_name":"My Blog","article_published_time":"2026-11-18T10:39:00+00:00","og_image":[{"url":"https:\/\/mmx.prnasia.com\/media\/MS2038583\/20261118012432EST_image_3.jpg?id=OA3087316&amp;p=medium600","type":"","width":"","height":""}],"author":"admin_tony","twitter_card":"summary_large_image","twitter_misc":{"Written by":"admin_tony","Est. reading time":"4 minutes"},"schema":{"@context":"https:\/\/schema.org","@graph":[{"@type":"Article","@id":"https:\/\/transmediavictoria.net.au\/?p=49375#article","isPartOf":{"@id":"https:\/\/transmediavictoria.net.au\/?p=49375"},"author":{"name":"admin_tony","@id":"https:\/\/transmediavictoria.net.au\/#\/schema\/person\/5b51d407d652c0aa802bea269ea5d7e7"},"headline":"Einsia AI Launches Visual Cipher Bench to Evaluate Long-Horizon Visual Reasoning in AI Agents","datePublished":"2026-11-18T10:39:00+00:00","mainEntityOfPage":{"@id":"https:\/\/transmediavictoria.net.au\/?p=49375"},"wordCount":768,"image":{"@id":"https:\/\/transmediavictoria.net.au\/?p=49375#primaryimage"},"thumbnailUrl":"https:\/\/transmediavictoria.net.au\/wp-content\/uploads\/2026\/11\/einsia-ai-launches-visual-cipher-bench-to-evaluate-long-horizon-visual-reasoning-in-ai-agents.jpg","articleSection":["Business News"],"inLanguage":"en-US"},{"@type":"WebPage","@id":"https:\/\/transmediavictoria.net.au\/?p=49375","url":"https:\/\/transmediavictoria.net.au\/?p=49375","name":"Einsia AI Launches Visual Cipher Bench to Evaluate Long-Horizon Visual Reasoning in AI Agents - My Blog","isPartOf":{"@id":"https:\/\/transmediavictoria.net.au\/#website"},"primaryImageOfPage":{"@id":"https:\/\/transmediavictoria.net.au\/?p=49375#primaryimage"},"image":{"@id":"https:\/\/transmediavictoria.net.au\/?p=49375#primaryimage"},"thumbnailUrl":"https:\/\/transmediavictoria.net.au\/wp-content\/uploads\/2026\/11\/einsia-ai-launches-visual-cipher-bench-to-evaluate-long-horizon-visual-reasoning-in-ai-agents.jpg","datePublished":"2026-11-18T10:39:00+00:00","author":{"@id":"https:\/\/transmediavictoria.net.au\/#\/schema\/person\/5b51d407d652c0aa802bea269ea5d7e7"},"breadcrumb":{"@id":"https:\/\/transmediavictoria.net.au\/?p=49375#breadcrumb"},"inLanguage":"en-US","potentialAction":[{"@type":"ReadAction","target":["https:\/\/transmediavictoria.net.au\/?p=49375"]}]},{"@type":"ImageObject","inLanguage":"en-US","@id":"https:\/\/transmediavictoria.net.au\/?p=49375#primaryimage","url":"https:\/\/transmediavictoria.net.au\/wp-content\/uploads\/2026\/11\/einsia-ai-launches-visual-cipher-bench-to-evaluate-long-horizon-visual-reasoning-in-ai-agents.jpg","contentUrl":"https:\/\/transmediavictoria.net.au\/wp-content\/uploads\/2026\/11\/einsia-ai-launches-visual-cipher-bench-to-evaluate-long-horizon-visual-reasoning-in-ai-agents.jpg","width":599,"height":268},{"@type":"BreadcrumbList","@id":"https:\/\/transmediavictoria.net.au\/?p=49375#breadcrumb","itemListElement":[{"@type":"ListItem","position":1,"name":"Home","item":"https:\/\/transmediavictoria.net.au\/"},{"@type":"ListItem","position":2,"name":"Einsia AI Launches Visual Cipher Bench to Evaluate Long-Horizon Visual Reasoning in AI Agents"}]},{"@type":"WebSite","@id":"https:\/\/transmediavictoria.net.au\/#website","url":"https:\/\/transmediavictoria.net.au\/","name":"My Blog","description":"My WordPress Blog","potentialAction":[{"@type":"SearchAction","target":{"@type":"EntryPoint","urlTemplate":"https:\/\/transmediavictoria.net.au\/?s={search_term_string}"},"query-input":{"@type":"PropertyValueSpecification","valueRequired":true,"valueName":"search_term_string"}}],"inLanguage":"en-US"},{"@type":"Person","@id":"https:\/\/transmediavictoria.net.au\/#\/schema\/person\/5b51d407d652c0aa802bea269ea5d7e7","name":"admin_tony","image":{"@type":"ImageObject","inLanguage":"en-US","@id":"https:\/\/secure.gravatar.com\/avatar\/102836fe5f9cd6c64e96e90e7341dbc815b7d12726a85708326dc0db40afd906?s=96&d=mm&r=g","url":"https:\/\/secure.gravatar.com\/avatar\/102836fe5f9cd6c64e96e90e7341dbc815b7d12726a85708326dc0db40afd906?s=96&d=mm&r=g","contentUrl":"https:\/\/secure.gravatar.com\/avatar\/102836fe5f9cd6c64e96e90e7341dbc815b7d12726a85708326dc0db40afd906?s=96&d=mm&r=g","caption":"admin_tony"},"url":"https:\/\/transmediavictoria.net.au\/?author=8"}]}},"_links":{"self":[{"href":"https:\/\/transmediavictoria.net.au\/index.php?rest_route=\/wp\/v2\/posts\/49375","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/transmediavictoria.net.au\/index.php?rest_route=\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/transmediavictoria.net.au\/index.php?rest_route=\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/transmediavictoria.net.au\/index.php?rest_route=\/wp\/v2\/users\/8"}],"replies":[{"embeddable":true,"href":"https:\/\/transmediavictoria.net.au\/index.php?rest_route=%2Fwp%2Fv2%2Fcomments&post=49375"}],"version-history":[{"count":0,"href":"https:\/\/transmediavictoria.net.au\/index.php?rest_route=\/wp\/v2\/posts\/49375\/revisions"}],"wp:featuredmedia":[{"embeddable":true,"href":"https:\/\/transmediavictoria.net.au\/index.php?rest_route=\/wp\/v2\/media\/49376"}],"wp:attachment":[{"href":"https:\/\/transmediavictoria.net.au\/index.php?rest_route=%2Fwp%2Fv2%2Fmedia&parent=49375"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/transmediavictoria.net.au\/index.php?rest_route=%2Fwp%2Fv2%2Fcategories&post=49375"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/transmediavictoria.net.au\/index.php?rest_route=%2Fwp%2Fv2%2Ftags&post=49375"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}