Research map / Image generation / Text-to-image diffusion
PosterAgent: Agentic Poster Generation via Stage-Aware Reinforcement Learning
Zhuocheng Yu, Feng Zhang, Sujian Li, Kai Jia
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 iterative self-refinement turns (ablation) | PosterAgent (Qwen3-VL-8B) · PosterAgent-Test | Overall score: 6.80 -> 7.41 | Improved |
| use stage-aware reward (SARL) instead of outcome-only reward (ablation) | PosterAgent · PosterAgent-Test | Overall score: 6.95 -> 7.41 | Improved |
| add renderer execution warnings as textual feedback (ablation) | PosterAgent · PosterAgent-Test | Overall score: 7.16 -> 7.41 | Improved |
| increase refinement steps from 3 to 4 (ablation) | PosterAgent · PosterAgent-Test | Overall score: 7.41 -> 7.39 | No clear effect |
| increase refinement steps from 0 to 3 (ablation) | PosterAgent · PosterAgent-Test | Overall score: 6.80 -> 7.41 | Improved |
| add RL after cold-start SFT | PosterAgent · PosterAgent-Test | Overall score: 6.76 -> 7.41 | Improved |
| add RL after cold-start SFT | PosterAgent · External out-of-domain benchmark | Overall score: 6.87 -> 7.66 | Improved |
| fine-tune Qwen3-VL-8B into PosterAgent | Qwen3-VL-8B · PosterAgent-Test | Overall score: 5.08 -> 7.41 | Improved |
| use editable JSON design agent instead of text-to-image generation | PosterAgent · PosterAgent-Test | Text quality: 5.04 -> 7.32 | Improved |
| PosterAgent vs GPT-5 | PosterAgent · PosterAgent-Test | Overall score: 7.43 vs 7.41 | No clear effect |
Closest papers by meaning
- Presenting a Paper is an Art: Self-Improvement Aesthetic Agents for Academic Presentations (ICLR 2026 · similarity 0.87)
- Generation Navigator: A State-Aware Agentic Framework for Image Generation (NeurIPS 2026 · similarity 0.86)
- Paper2Poster: Towards Multimodal Poster Automation from Scientific Papers (NeurIPS 2025 · similarity 0.84)
- POSTA: A Go-to Framework for Customized Artistic Poster Generation (CVPR 2025 · similarity 0.84)
- Agentic Design Review System (AAAI 2026 · similarity 0.84)
- Graphic Design with Large Multimodal Model (AAAI 2025 · similarity 0.83)
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.
