IV-CoT: Implicit Visual Chain-of-Thought for Structure-Aware Text-to-Image Generation
Summary
IV-CoT decomposes visual conditioning into structural and semantic cascades for improved structure-aware image generation, using training-only sketch supervision to guide structural queries. It achieves state-of-the-art results on GenEval and T2I-CompBench.
View Cached Full Text
Cached at: 06/25/26, 05:17 AM
Paper page - IV-CoT: Implicit Visual Chain-of-Thought for Structure-Aware Text-to-Image Generation
Source: https://huggingface.co/papers/2606.24849 Authors:
,
,
,
,
,
,
,
,
,
,
,
Abstract
Implicit Visual Chain-of-Thought decomposes visual conditioning into structural and semantic cascades for improved structure-aware image generation with sketch supervision.
Unifiedmulti-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 ofstructural planningandappearance renderingwithin a single conditioning stream. To address this issue, we propose Implicit VisualChain-of-Thought(IV-CoT), alatent visual reasoningframework for query-conditioned image generation. IV-CoT decomposes thevisual conditioningqueries 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-onlysketch 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 instructure-aware generation.
View arXiv pageView PDFAdd to collection
Get this paper in your agent:
hf papers read 2606\.24849
Don’t have the latest CLI?curl \-LsSf https://hf\.co/cli/install\.sh \| bash
Models citing this paper0
No model linking this paper
Cite arxiv.org/abs/2606.24849 in a model README.md to link it from this page.
Datasets citing this paper0
No dataset linking this paper
Cite arxiv.org/abs/2606.24849 in a dataset README.md to link it from this page.
Spaces citing this paper0
No Space linking this paper
Cite arxiv.org/abs/2606.24849 in a Space README.md to link it from this page.
Collections including this paper0
No Collection including this paper
Add this paper to acollectionto link it from this page.
Similar Articles
Beyond Visual CoT: Internalized Visual Thinking for Proactive Video Reasoning
The paper introduces Internalized Visual Thinking (IVT), a post-training framework that trains multimodal models to predict future frame embeddings, enabling direct answer generation without synthesizing intermediate images, thereby reducing latency over 5x for proactive video reasoning.
Unlocking Complex Visual Generation via Closed-Loop Verified Reasoning
Introduces CLVR (Closed-Loop Visual Reasoning), a framework that reformulates text-to-image generation from a single-step process into a closed-loop, multi-step visual reasoning approach using a VLM controller and diffusion models, achieving improved performance on compositional prompts.
J-CoT: Chain-of-Thought in J-Space
This paper introduces J-CoT, a recurrent reasoning framework that uses vocabulary-indexed coefficients (J-thoughts) as intermediate interfaces, enabling improved reasoning performance on mathematical, scientific, coding, and path-reasoning tasks without requiring full verbalization or dense hidden state recurrence.
CoVA-SFT: A Large-Scale Dataset for Chain of Visual Abstractions
CoVA-SFT is a large-scale dataset and benchmark designed to train multimodal language models in chain of visual abstractions, improving performance over baselines in visual reasoning tasks.
ViCo: Visual-oriented Coding with Self-Reflection for Chart Replication
ViCo is a training framework that uses iterative self-reflection to improve AI-generated chart replication, achieving performance close to proprietary LLMs on academic benchmarks.