CVPR Findings2026
DraCo: Draft as CoT for Text-to-Image Preview and Rare Concept Generation
Dongzhi Jiang, Renrui Zhang, Haodong Li, Zhuofan Zong, Ziyu Guo, Jun He, Claire Guo, Junyan Ye, Rongyao Fang, Weijia Li, Rui Liu, Hongsheng Li
Abstract
Recent unified multimodal large language models (MLLMs) have shown impressive capabilities, incorporating chainof-thought (CoT) reasoning for enhanced text-to-image generation. However, existing approaches remain limited, either treating the model merely as a standalone generator or relying on abstract textual planning. To this end, we propose Draft-as-CoT (DraCo), a novel interleaved reasoning paradigm that fully leverages both textual and visual contents in CoT for better planning and verification. Our method first generates a low-resolution draft image as preview, providing more concrete and structural visual planning and guidance. Then, we employ the model's inherent understanding capability to verify potential semantic misalignments between the draft and input prompt, and performs refinement through selective corrections with super-resolution. In this way, our approach addresses two fundamental challenges: the coarse-grained nature of textual planning and the difficulty in generating rare attribute combinations. To support training, we curate DraCo-240K, aiming to enhance three atomic capabilities spanning general correction, instance manipulation, and layout reorganization. Supported by DraCo-CFG, a specialized classifier-free guidance (CFG) strategy for interleaved reasoning, DraCo achieves a tremendous increase on GenEval (+8%), Imagine-Bench (+0.91), and GenEval++ (+3%), significantly outperforming direct generation and other generation methods empowered by CoT.
Ask Lune
Know this paper in minutes, not an afternoon.
Lune can read every page of it and the top-tier papers that cite it, and gives you its claims, its evidence and its limits, each one cited to the paper it comes from. No skimming, no AI slop.
2 top-tier papers have cited it since. Know what they found before your reviewers do.
Lune can draw on
the full paper · the top-tier papers that cite it · related work
Free to start. No credit card required.
Lineage
- SDXL: Improving Latent Diffusion Models for High-Resolution Image Synthesis cites High-Resolution Image Synthesis with Latent Diffusion Models
- Scaling Rectified Flow Transformers for High-Resolution Image Synthesis cites High-Resolution Image Synthesis with Latent Diffusion Models
- Scaling Rectified Flow Transformers for High-Resolution Image Synthesis cites SDXL: Improving Latent Diffusion Models for High-Resolution Image Synthesis
- CoF-T2I: Video Models as Pure Visual Reasoners for Text-to-Image Generation cites SDXL: Improving Latent Diffusion Models for High-Resolution Image Synthesis
- CoF-T2I: Video Models as Pure Visual Reasoners for Text-to-Image Generation cites Chain-of-Thought Prompting Elicits Reasoning in Large Language Models
- CoF-T2I: Video Models as Pure Visual Reasoners for Text-to-Image Generation cites T2I-R1: Reinforcing Image Generation with Collaborative Semantic-level and Token-level CoT
- T2I-R1: Reinforcing Image Generation with Collaborative Semantic-level and Token-level CoT cites High-Resolution Image Synthesis with Latent Diffusion Models
- T2I-R1: Reinforcing Image Generation with Collaborative Semantic-level and Token-level CoT cites SDXL: Improving Latent Diffusion Models for High-Resolution Image Synthesis
- T2I-R1: Reinforcing Image Generation with Collaborative Semantic-level and Token-level CoT cites Chain-of-Thought Prompting Elicits Reasoning in Large Language Models
Cited by top-tier papers2
- T2I-R1: Reinforcing Image Generation with Collaborative Semantic-level and Token-level CoTDongzhi Jiang, Ziyu Guo, Renrui Zhang, Zhuofan Zong et al.NeurIPS 2025 · 181 citations
- CoF-T2I: Video Models as Pure Visual Reasoners for Text-to-Image GenerationChengzhuo Tong, Chang Mingkun, Shenglong Zhang, Yuran Wang et al.ICML 2026 · 7 citations
Builds on31
- Chain-of-Thought Prompting Elicits Reasoning in Large Language ModelsJason Wei, Xuezhi Wang, Dale Schuurmans, Maarten Bosma et al.NeurIPS 2022 · 22,562 citations
- High-Resolution Image Synthesis with Latent Diffusion ModelsRobin Rombach, Andreas Blattmann, Dominik Lorenz, Patrick Esser et al.CVPR 2022 · 13,123 citations
- SDXL: Improving Latent Diffusion Models for High-Resolution Image SynthesisDustin Podell, Zion English, Kyle Lacey, Andreas Blattmann et al.ICLR 2024 · 4,569 citations
- Scaling Rectified Flow Transformers for High-Resolution Image SynthesisPatrick Esser, Sumith Kulal, Andreas Blattmann, Rahim Entezari et al.ICML 2024 · 3,620 citations
- LongRoPE: Extending LLM Context Window Beyond 2 Million TokensYiran Ding, Li Lyna Zhang, Chengruidong Zhang, Yuanyuan Xu et al.ICML 2024 · 316 citations
Related papers
- GoT: Unleashing Reasoning Capability of MLLM for Visual Generation and EditingRongyao Fang, Chengqi Duan, Kun Wang, Linjiang Huang et al.NeurIPS 2025 · 5 citations
- UniGen: Enhanced Training & Test-Time Strategies for Unified Multimodal Understanding and GenerationRui Tian, Mingfei Gao, Mingze Xu, Jiaming Hu et al.NeurIPS 2025 · 35 citations
- CoT-lized Diffusion: Let's Reinforce T2I Generation Step-by-stepZheyuan Liu, Munan Ning, Qihui Zhang, Shuo Yang et al.NeurIPS 2025 · 9 citations
- Enhance Multimodal Consistency and Coherence for Text-Image Plan GenerationXiaoxin Lu, Ranran Haoran Zhang, Yusen Zhang, Rui ZhangACL Findings 2025
- Vinci: Deep Thinking in Text-to-Image Generation using Unified Model with Reinforcement LearningWang Lin, Wentao Hu, Liyu Jia, Kaihang Pan et al.NeurIPS 2025
