README.md

---
title: RAP — Paper to Poster
emoji: 📰
colorFrom: purple
colorTo: blue
sdk: gradio
sdk_version: 5.50.0
python_version: 3.12
app_file: app.py
pinned: false
startup_duration_timeout: 1h
short_description: Turn a research paper PDF into an editable academic poster
---

# RAP · Paper to Poster

[![Paper](https://img.shields.io/badge/Paper-arXiv%3A2610.23230-B31B1B?logo=arxiv&logoColor=fff)](https://arxiv.org/abs/2610.23230)
[![Project Page](https://img.shields.io/badge/Project-Page-8A2BE2?logo=githubpages&logoColor=fff)](https://reconstruct-a-poster.github.io/)
[![Code](https://img.shields.io/badge/Code-GitHub-181717?logo=github&logoColor=fff)](https://github.com/VincentHancoder/RAP)
[![Dataset](https://img.shields.io/badge/Dataset-RAP--25K-FFD21E?logo=huggingface&logoColor=000)](https://huggingface.co/datasets/VincentHancoder/RAP)
[![Model](https://img.shields.io/badge/Model-RAP--Qwen3.8--27B--RL-FF9D00?logo=huggingface&logoColor=000)](https://huggingface.co/VincentHancoder/RAP-Qwen3.8-27B-RL)

Upload a research paper PDF and get an editable academic poster ([paper](https://arxiv.org/abs/2610.23230), [project page](https://reconstruct-a-poster.github.io/), [code](https://github.com/VincentHancoder/RAP)). A 27B vision-language model
(Qwen3.8-27B, SFT + GRPO on real conference posters) reads the paper, selects figures and tables,
and writes the poster as standalone HTML; the Space renders it to PNG.

- Model: RL checkpoint-240, BF16, no quantization, 16K context.
- The model runs in BF16 on a full 96 GB ZeroGPU slot (`xlarge`, counted 2x against quota). A poster
  takes about 15–25 minutes of ZeroGPU quota; PRO accounts have 40 min/day, free accounts 5 min and
  anonymous visitors 2 min, so sign in with a PRO account (on huggingface.co, not the bare hf.space URL)
  to finish a poster. The demo asks for a 10-minute GPU step first and falls back to shorter steps.
- API: call `/analyze` once (PDF + settings), then `/step` repeatedly until the status starts with
  `Done`, `Paused` or `Stopped`; each step is its own request so the ZeroGPU token stays valid.
- The example cards are five cases of the RAP-150 benchmark. Picking one loads the paper, the exact instruction and
  layout specification used in the paper's evaluation and the canvas size of the reference poster; **Generate poster**
  then runs the model on the very prompt we evaluated, so the result should match the shown poster. Editing any field
  switches back to the regular pipeline (your PDF is parsed and a layout specification is derived automatically).
- Research demo: scanned PDFs are not supported, uploads are temporary, and every statement,
  number and caption should be checked before a poster is used.

Third-party components (PDF-Extract-Kit LayoutLMv3 detector and the training preprocessing
modules) are listed in [THIRD_PARTY_NOTICES.md](THIRD_PARTY_NOTICES.md).

## Citation

```bibtex
@misc{han2026rapdatasetpapertopostergeneration,
      title={RAP: A Dataset for Paper-to-Poster Generation via Declarative Markup from Human-Crafted Exemplars},
      author={Haonan Han and Yukang Lin and Minliang Lin and Haoxian Tan and Jun-Yan He and Rui Yang and Zhi-Qi Cheng and Jie Hu and Guanglu Wan and Xiu Li},
      year={2026},
      eprint={2610.23230},
      archivePrefix={arXiv},
      primaryClass={cs.CV},
      url={https://arxiv.org/abs/2610.23230},
}
```