cs.CVJun 23, 2026

IV-CoT: Implicit Visual Chain-of-Thought for Structure-Aware Text-to-Image Generation

Authors: Zixuan LiHaokun LinYicheng XiaoZhiwei LiXinyang SongZelong ZhengYong HeHeng Yao+5 more

Organizations: NLPR, Institute of Automation, Chinese Academy of Sciences · The University of Hong Kong · Ant Group

Abstract

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.

Explore similar work

CardsList
  1. Tree-of-Thoughts Reasoning for Text-to-Image In-Context Learning

    Jul 8, 2026Stepanida Alekseeva, Jenifer Kalafatovich, Seong-Whan LeeText-To-ImageIn-Context Learning