IV-CoT introduces an elegant implicit visual reasoning mechanism for text-to-image generation, separating structural planning from semantic rendering to improve compositional fidelity in counts, spatial relations, and layouts without requiring explicit intermediate CoT outputs at inference.</p>\n","updatedAt":"2026-06-25T02:01:33.921Z","author":{"_id":"6841378dcb9938c7aed795be","avatarUrl":"https://cdn-avatars.huggingface.co/v1/production/uploads/no-auth/WNI8gQVHdWyt0OvmRINDp.png","fullname":"Haokun Lin","name":"Felix1023","type":"user","isPro":false,"isHf":false,"isHfAdmin":false,"isMod":false,"followerCount":4,"isUserFollowing":false}},"numEdits":0,"identifiedLanguage":{"language":"en","probability":0.7334791421890259},"editors":["Felix1023"],"editorAvatarUrls":["https://cdn-avatars.huggingface.co/v1/production/uploads/no-auth/WNI8gQVHdWyt0OvmRINDp.png"],"reactions":[],"isReport":false}}],"primaryEmailConfirmed":false,"paper":{"id":"2606.24849","authors":[{"_id":"6a3c8b89f3facdb67e9ff03e","name":"Zixuan Li","hidden":false},{"_id":"6a3c8b89f3facdb67e9ff03f","name":"Haokun Lin","hidden":false},{"_id":"6a3c8b89f3facdb67e9ff040","name":"Yicheng Xiao","hidden":false},{"_id":"6a3c8b89f3facdb67e9ff041","name":"Zhiwei Li","hidden":false},{"_id":"6a3c8b89f3facdb67e9ff042","name":"Xinyang Song","hidden":false},{"_id":"6a3c8b89f3facdb67e9ff043","name":"Zelong Zheng","hidden":false},{"_id":"6a3c8b89f3facdb67e9ff044","name":"Yong He","hidden":false},{"_id":"6a3c8b89f3facdb67e9ff045","name":"Heng Yao","hidden":false},{"_id":"6a3c8b89f3facdb67e9ff046","name":"Ke Ding","hidden":false},{"_id":"6a3c8b89f3facdb67e9ff047","name":"Chao Yu","hidden":false},{"_id":"6a3c8b89f3facdb67e9ff048","name":"Chuan Yuan","hidden":false},{"_id":"6a3c8b89f3facdb67e9ff049","name":"Qi Li","hidden":false},{"_id":"6a3c8b89f3facdb67e9ff04a","name":"Zhenan Sun","hidden":false}],"publishedAt":"2026-06-23T00:00:00.000Z","submittedOnDailyAt":"2026-06-25T00:00:00.000Z","title":"IV-CoT: Implicit Visual Chain-of-Thought for Structure-Aware Text-to-Image Generation","submittedOnDailyBy":{"_id":"6841378dcb9938c7aed795be","avatarUrl":"https://cdn-avatars.huggingface.co/v1/production/uploads/no-auth/WNI8gQVHdWyt0OvmRINDp.png","isPro":false,"fullname":"Haokun Lin","user":"Felix1023","type":"user","name":"Felix1023"},"summary":"Unified multi-modal large language models (MLLMs) have achieved strong text-to-image generation quality, but still struggle with structure-aware prompt following, where object counts, spatial relations, attribute bindings, and coarse layouts must be preserved. We attribute this limitation in part to the entanglement of structural planning and appearance rendering within a single conditioning stream. To address this issue, we propose Implicit Visual Chain-of-Thought (IV-CoT), a latent visual reasoning framework for query-conditioned image generation. IV-CoT decomposes the visual conditioning queries into a structural-to-semantic cascade, where structural queries first form a latent visual plan and semantic queries then render appearance conditioned on this plan. To guide the structural queries, we introduce training-only sketch supervision, which encourages them to capture structure from sketches without requiring sketch extraction or intermediate decoding at inference time. IV-CoT performs implicit CoT reasoning in a single forward pass and achieves superior results on GenEval and T2I-CompBench. Visualizations and analyses demonstrate that the learned structural and semantic queries play complementary roles in structure-aware generation.","upvotes":11,"discussionId":"6a3c8b89f3facdb67e9ff04b","ai_summary":"Implicit Visual Chain-of-Thought decomposes visual conditioning into structural and semantic cascades for improved structure-aware image generation with sketch supervision.","ai_keywords":["multi-modal large language models","visual conditioning","structural planning","appearance rendering","latent visual reasoning","chain-of-thought","sketch supervision","visual planning","structure-aware generation"],"ai_summary_model":"Qwen/Qwen2.5-Coder-32B-Instruct","organization":{"_id":"67c1d682826160b28f778510","name":"antgroup","fullname":"Ant Group","avatar":"https://cdn-avatars.huggingface.co/v1/production/uploads/662e1f9da266499277937d33/7VcPHdLSGlged3ixK1dys.jpeg"}},"canReadDatabase":false,"canManagePapers":false,"canSubmit":false,"hasHfLevelAccess":false,"upvoted":false,"upvoters":[{"_id":"6841378dcb9938c7aed795be","avatarUrl":"https://cdn-avatars.huggingface.co/v1/production/uploads/no-auth/WNI8gQVHdWyt0OvmRINDp.png","isPro":false,"fullname":"Haokun Lin","user":"Felix1023","type":"user"},{"_id":"68cba1e2f02ac083a52cbbf0","avatarUrl":"/avatars/4459a855cc844365c3ff4195bf680238.svg","isPro":false,"fullname":"lzx","user":"UGuess23333","type":"user"},{"_id":"641fb3733c49fc33cedb4429","avatarUrl":"/avatars/3577bca726282ed4b26009f3a2d5e87d.svg","isPro":false,"fullname":"Kaijie Zhu","user":"March07","type":"user"},{"_id":"66ac46c1308c1f8ae90d50bb","avatarUrl":"/avatars/0a7caeb33b513a04c1c44714ac8a2854.svg","isPro":false,"fullname":"wang wei","user":"davidwang39","type":"user"},{"_id":"65950d64a02954c982f1938c","avatarUrl":"/avatars/ddbdb2b1bafbbd91f19ab3ba247c2152.svg","isPro":false,"fullname":"Yuxuan Hu","user":"huyx55","type":"user"},{"_id":"69a29961f36d2a7fb93f2e77","avatarUrl":"/avatars/7a271e0aa7a38400298043e601ef34c7.svg","isPro":false,"fullname":"rumn","user":"Rumny","type":"user"},{"_id":"68c7aa9f740391d5f3a7f01c","avatarUrl":"/avatars/ec9a147240c0deb5fe143e024a8e6be6.svg","isPro":false,"fullname":"tom","user":"dotoary","type":"user"},{"_id":"66b09d0a7cb12daaffcba47e","avatarUrl":"/avatars/ed25d0a3d816dff4e86e7c59989f7277.svg","isPro":false,"fullname":"Matcha","user":"MatchaLwc","type":"user"},{"_id":"67c14b84a261b07d0f8dadc5","avatarUrl":"https://cdn-avatars.huggingface.co/v1/production/uploads/no-auth/YVfwrVAXFWyXV5w9NyzJr.png","isPro":false,"fullname":"Jiexiang","user":"Yecole525","type":"user"},{"_id":"633fd532d4fc96dd3a04ae89","avatarUrl":"/avatars/f373772e5602320f685173c94507ba42.svg","isPro":false,"fullname":"sunxin","user":"unixin","type":"user"},{"_id":"6860f511b8c6f8138ab39181","avatarUrl":"/avatars/a7b8fc921f9727e11dfc601b4b010b26.svg","isPro":false,"fullname":"Jian Chen","user":"cccccj03","type":"user"}],"acceptLanguages":["en"],"dailyPaperRank":0,"organization":{"_id":"67c1d682826160b28f778510","name":"antgroup","fullname":"Ant Group","avatar":"https://cdn-avatars.huggingface.co/v1/production/uploads/662e1f9da266499277937d33/7VcPHdLSGlged3ixK1dys.jpeg"},"markdownContentUrl":"https://huggingface.co/buckets/huggingchat/papers-content/resolve/2606/2606.24849.md","query":{}}">

Get this paper in your agent:

Don't have the latest CLI?

Cite arxiv.org/abs/2606.24849 in a model README.md to link it from this page.

Cite arxiv.org/abs/2606.24849 in a dataset README.md to link it from this page.

Cite arxiv.org/abs/2606.24849 in a Space README.md to link it from this page.

Add this paper to a collection to link it from this page.