--- title: Beyond Visual CoT: Internalized Visual Thinking for... description: Apple researchers propose Internalized Visual Thinking (IVT), a post-training framework for multimodal LLMs that predicts latent representations of future... canonical: https://daily.dev/posts/beyond-visual-cot-internalized-visual-thinking-for-proactive-video-reasoning-ryknzthep twitter:card: summary_large_image twitter:site: @dailydotdev og:type: website og:site_name: daily.dev og:title: Beyond Visual CoT: Internalized Visual Thinking for Proactive Video Reasoning | daily.dev og:description: Apple researchers propose Internalized Visual Thinking (IVT), a post-training framework for multimodal LLMs that predicts latent representations of future... og:url: https://daily.dev/posts/beyond-visual-cot-internalized-visual-thinking-for-proactive-video-reasoning-ryknzthep og:image: https://api.daily.dev/og/posts/ryknZThep.png og:image:alt: Beyond Visual CoT: Internalized Visual Thinking for Proactive Video Reasoning og:image:width: 1200 og:image:height: 630 og:locale: en --- > ## Documentation Index > Fetch the complete documentation index at: https://daily.dev/llms.txt > Use this file to discover all available pages before exploring further. # Beyond Visual CoT: Internalized Visual Thinking for Proactive Video Reasoning **[Apple Machine Learning Research](https://daily.dev/sources/mlapple)** · 2 min read · 1 upvotes · 0 comments ## Summary Apple researchers propose Internalized Visual Thinking (IVT), a post-training framework for multimodal LLMs that predicts latent representations of future video frames alongside textual answers, instead of generating explicit intermediate images as Visual Chain-of-Thought does. This lets models reason proactively about video without the inference overhead of synthesizing or re-encoding frames. Across six evaluation settings, IVT outperforms text-only post-training and matches or beats Visual CoT while cutting end-to-end latency by more than 5x, suggesting pixel-space generation may be unnecessary for efficient video reasoning. ## Full article daily.dev links to this article rather than hosting it. Read it at the original source: ## Questions this post answers ### What is Internalized Visual Thinking (IVT) in multimodal video reasoning models? Internalized Visual Thinking is a post-training framework that jointly optimizes textual prediction and next-embedding prediction over unlabeled videos, so a model predicts latent representations of future frames along with the textual answer during training. At inference, the model generates answers directly without synthesizing or re-encoding future frames, avoiding the overhead of explicit visual chain-of-thought. _Teams building efficient video-reasoning models can track approaches like this on daily.dev._ ### How much faster is IVT compared to Visual Chain-of-Thought (Visual CoT) for video reasoning? IVT reduces end-to-end latency by more than 5x compared to Visual CoT while achieving comparable or better performance across six evaluation settings. This is because IVT avoids generating and re-encoding intermediate reasoning images at inference, instead relying on latent future-frame predictions learned during training. _Anyone comparing reasoning approaches for video models can follow efficiency benchmarks like this on daily.dev._ ## Similar posts on daily.dev - [How visual learning happens in the brain](https://daily.dev/posts/how-visual-learning-happens-in-the-brain-yvbujwok8) · MIT News · 54 upvotes · 3 comments - [Visualizing Vision-Language Models](https://daily.dev/posts/visualizing-vision-language-models-3etxgb4cm) · DigitalOcean Community · 0 upvotes · 0 comments - [How Vision Language Models Are Trained from “Scratch”](https://daily.dev/posts/how-vision-language-models-are-trained-from-scratch--cptxwc95l) · Towards Data Science · 0 upvotes · 0 comments - [Multimodal LLMs Basics: How LLMs Process Text, Images, Audio & Videos](https://daily.dev/posts/multimodal-llms-basics-how-llms-process-text-images-audio-videos-fw0j9el9k) · ByteByteGo · 5 upvotes · 0 comments --- Tags: [#apple](https://daily.dev/tags/apple), [#prompt-engineering](https://daily.dev/tags/prompt-engineering), [#world-models](https://daily.dev/tags/world-models) [View this post on daily.dev](https://daily.dev/posts/beyond-visual-cot-internalized-visual-thinking-for-proactive-video-reasoning-ryknzthep) ```json {"@context":"https://schema.org","@graph":[{"@type":"Organization","@id":"https://daily.dev/#organization","name":"daily.dev","url":"https://daily.dev","logo":{"@type":"ImageObject","url":"https://daily.dev/apple-touch-icon.png","width":180,"height":180},"sameAs":["https://twitter.com/dailydotdev","https://github.com/dailydotdev","https://www.linkedin.com/company/daily-dev-ltd"]},{"@type":"WebSite","@id":"https://daily.dev/#website","url":"https://daily.dev","name":"daily.dev","publisher":{"@id":"https://daily.dev/#organization"},"potentialAction":{"@type":"SearchAction","target":{"@type":"EntryPoint","urlTemplate":"https://daily.dev/search?q={search_term_string}"},"query-input":"required name=search_term_string"}}]} {"@context":"https://schema.org","@type":"TechArticle","headline":"Beyond Visual CoT: Internalized Visual Thinking for Proactive Video Reasoning","url":"https://daily.dev/posts/beyond-visual-cot-internalized-visual-thinking-for-proactive-video-reasoning-ryknzthep","mainEntityOfPage":{"@type":"WebPage","@id":"https://daily.dev/posts/beyond-visual-cot-internalized-visual-thinking-for-proactive-video-reasoning-ryknzthep"},"datePublished":"2026-08-24T15:42:24.652Z","dateModified":"2026-09-13T20:10:28.192Z","description":"Apple researchers propose Internalized Visual Thinking (IVT), a post-training framework for multimodal LLMs that predicts latent representations of future...","image":"https://media.daily.dev/image/upload/f_auto,q_auto/v1/posts/8dda20dea5f9a8a0e25a1b7bb36a1a20?_a=AQAEuop","thumbnailUrl":"https://media.daily.dev/image/upload/f_auto,q_auto/v1/posts/8dda20dea5f9a8a0e25a1b7bb36a1a20?_a=AQAEuop","isAccessibleForFree":true,"articleSection":"Apple Machine Learning Research","inLanguage":"en","publisher":{"@type":"Organization","name":"daily.dev","url":"https://daily.dev","logo":{"@type":"ImageObject","url":"https://daily.dev/apple-touch-icon.png","width":180,"height":180}},"author":{"@type":"Organization","name":"Apple Machine Learning Research","logo":"https://media.daily.dev/image/upload/s--mD-1uSe0--/f_auto,q_auto/v1785654191/logos/mlapple","url":"https://daily.dev/sources/mlapple"},"commentCount":0,"discussionUrl":"https://daily.dev/posts/beyond-visual-cot-internalized-visual-thinking-for-proactive-video-reasoning-ryknzthep","interactionStatistic":[{"@type":"InteractionCounter","interactionType":{"@type":"LikeAction"},"userInteractionCount":1},{"@type":"InteractionCounter","interactionType":{"@type":"CommentAction"},"userInteractionCount":0}],"keywords":"apple,prompt-engineering,world-models","timeRequired":"PT2M"} {"@context":"https://schema.org","@type":"BreadcrumbList","itemListElement":[{"@type":"ListItem","position":1,"name":"Home","item":"https://daily.dev"},{"@type":"ListItem","position":2,"name":"Apple Machine Learning Research","item":"https://daily.dev/sources/mlapple"},{"@type":"ListItem","position":3,"name":"Beyond Visual CoT: Internalized Visual Thinking for Proactive Video Reasoning"}]} {"@context":"https://schema.org","@type":"FAQPage","@id":"https://daily.dev/posts/beyond-visual-cot-internalized-visual-thinking-for-proactive-video-reasoning-ryknzthep#faq","mainEntity":[{"@type":"Question","name":"What is Internalized Visual Thinking (IVT) in multimodal video reasoning models?","acceptedAnswer":{"@type":"Answer","text":"Internalized Visual Thinking is a post-training framework that jointly optimizes textual prediction and next-embedding prediction over unlabeled videos, so a model predicts latent representations of future frames along with the textual answer during training. At inference, the model generates answers directly without synthesizing or re-encoding future frames, avoiding the overhead of explicit visual chain-of-thought. Teams building efficient video-reasoning models can track approaches like this on daily.dev."}},{"@type":"Question","name":"How much faster is IVT compared to Visual Chain-of-Thought (Visual CoT) for video reasoning?","acceptedAnswer":{"@type":"Answer","text":"IVT reduces end-to-end latency by more than 5x compared to Visual CoT while achieving comparable or better performance across six evaluation settings. This is because IVT avoids generating and re-encoding intermediate reasoning images at inference, instead relying on latent future-frame predictions learned during training. Anyone comparing reasoning approaches for video models can follow efficiency benchmarks like this on daily.dev."}}]} ```