---
base_model:
- Qwen/Qwen3.5-9B
datasets:
- Forceless/UltraPresent
license: mit
library_name: transformers
pipeline_tag: text-generation
---

# DeepPresenter-9B

**Project**: [https://github.com/icip-cas/PPTAgent](https://github.com/icip-cas/PPTAgent)  
**Paper**: [DeepPresenter: Environment-Grounded Reflection for Agentic Presentation Generation](https://arxiv.org/abs/2602.22839)  

## Model Overview

**DeepPresenter-9B** is a 9B-parameter language model designed for **automatic presentation generation**. It serves as the core model in the **DeepPresenter** framework, enabling agentic workflows that generate structured slide presentations from natural language instructions. 

Unlike traditional agents that rely on fixed templates, DeepPresenter uses environment-grounded reflection. This allows the system to condition its generation process on perceptual artifact states (like rendered slides), enabling it to autonomously plan, render, and revise slides to correct presentation-specific issues during execution.

## Usage

DeepPresenter-9B is intended to be used with the **PPTAgent framework** for full presentation generation. You can use it via the CLI:

```bash
# Interactive configuration (first time)
uvx pptagent onboard

# Generate presentation
uvx pptagent generate "Single Page with Title: Hello World" -o hello.pptx
```

## Performance

![image](https://cdn-uploads.huggingface.co/production/uploads/64380d17819f3ab20d17595b/PZgEPAPir2ygRvxtqtXnx.png)

## Citation

If you find this work useful, please cite the paper:

```bibtex
@misc{zheng2026deeppresenterenvironmentgroundedreflectionagentic,
      title={DeepPresenter: Environment-Grounded Reflection for Agentic Presentation Generation},
      author={Hao Zheng and Guozhao Mo and Xinru Yan and Qianhao Yuan and Wenkai Zhang and Xuanang Chen and Yaojie Lu and Hongyu Lin and Xianpei Han and Le Sun},
      year={2026},
      eprint={2602.22839},
      archivePrefix={arXiv},
      primaryClass={cs.AI},
      url={https://arxiv.org/abs/2602.22839},
}
```