Research map / Image generation / Text-to-image diffusion
HiCoGen: Hierarchical Compositional Text-to-Image Generation in Diffusion Models via Reinforcement Learning
Hongji Yang, Yucheng Zhou, Wencheng Han, Runzhou Tao, Zhongying Qiu, Jianfei Yang, Jianbing Shen
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| add subject-level reward (ablation) | HiCoGen · HiCoPrompt | Acc_attribute: 0.6206 -> 0.7672 | Improved |
| add relationship-level reward (ablation) | HiCoGen · HiCoPrompt | Acc_relationship: 0.7689 -> 0.8202 | Improved |
| add global-level reward (ablation) | HiCoGen · HiCoPrompt | CLIP Score: 0.2670 -> 0.3192 | Improved |
| add global-level reward (ablation) | HiCoGen · HiCoPrompt | HPSv2: 0.2720 -> 0.3357 | Improved |
| add global-level reward (ablation) | HiCoGen · HiCoPrompt | Acc_attribute: 0.7743 -> 0.7672 | Hurt |
| add subject-level reward (ablation) | HiCoGen · HiCoPrompt | CLIP Score: 0.3208 -> 0.3192 | No clear effect |
| add global-level reward (ablation) | HiCoGen · HiCoPrompt | Acc_exist: 0.6828 -> 0.7128 | Improved |
| add relationship-level reward (ablation) | HiCoGen · HiCoPrompt | HPSv2: 0.3327 -> 0.3357 | No clear effect |
| use HiCoGen instead of Qwen-Image | HiCoGen · HiCoPrompt | Acc_exist: 0.6292 -> 0.7127 | Improved |
| use HiCoGen instead of FLUX.1-dev / Qwen-Image | HiCoGen · HiCoPrompt | HPSv2: 0.2974 -> 0.3357 | Improved |
| increase number of subjects in prompt from 1 to 3 | HiCoGen · HiCoPrompt | Acc_exist: 0.92 -> 0.41 | Hurt |
| use HiCoGen instead of Qwen-Image with 3 subjects | HiCoGen · HiCoPrompt (3 subjects) | Acc_relationship: 0.61 -> 0.71 | Improved |
Closest papers by meaning
- Sketch Then Paint: Hierarchical Reinforcement Learning for Diffusion Multi-Modal Large Language Models (NeurIPS 2026 · similarity 0.91)
- Learning to Sample Effective and Diverse Prompts for Text-to-Image Generation (CVPR 2025 · similarity 0.89)
- LLM Blueprint: Enabling Text-to-Image Generation with Complex and Detailed Prompts (ICLR 2024 · similarity 0.89)
- PromptCoT: Align Prompt Distribution via Adapted Chain-of-Thought (CVPR 2024 · similarity 0.89)
- Progressive Compositionality in Text-to-Image Generative Models (ICLR 2025 · similarity 0.89)
- Mastering Text-to-Image Diffusion: Recaptioning, Planning, and Generating with Multimodal LLMs (ICML 2024 · similarity 0.89)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.
