File size: 2,724 Bytes
e708893 b083592 e708893 04e55b7 e708893 704b3fe e708893 73ceab2 e708893 704b3fe 73ceab2 e708893 73ceab2 e708893 73ceab2 e708893 73ceab2 e708893 73ceab2 e708893 73ceab2 e708893 73ceab2 04e55b7 73ceab2 04e55b7 73ceab2 04e55b7 e708893 73ceab2 e708893 73ceab2 e708893 73ceab2 b083592 | 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 | ---
license: mit
language:
- en
tags:
- text-rewriting
- web
- generative-engine-optimization
- geo
- reinforcement-learning
- grpo
- qwen3
- transformers
- safetensors
library_name: transformers
pipeline_tag: text-generation
base_model: Qwen/Qwen3-1.7B
datasets:
- cx-cmu/GEO-Bench
---
# AutoGEO<sub>Mini</sub> (Qwen1.7B, GEO-Bench)
AutoGEO<sub>Mini</sub> (Qwen1.7B, GEO-Bench) is a GEO model designed to improve how web document is incorporated into answers generated by **LLM-based generative engines**.
The model rewrites a given document to better match the preferences of generative engines (e.g., GPT, Gemini, Claude), with the goal of increasing the document’s **visibility and coverage** in generated responses, while **preserving the original meaning and factual content**.
⚠️ This model is trained for the generative engine powered by `gemini-2.5-flash-lite` on dataset `GEO-Bench`. If you intend to use AutoGEO<sub>Mini</sub> with other types of generative engines or datasets, you must post-train `Qwen/Qwen3-1.7B` using [our code](https://github.com/cxcscmu/AutoGEO).
This model is part of the **AutoGEO** framework proposed in the paper
📄 **Paper:** ["What Generative Search Engines Like and How to Optimize Web Content Cooperatively"](https://arxiv.org/abs/2510.11438)
👥 **Authors:** Yujiang Wu*, Shanshan Zhong*, Yubin Kim, Chenyan Xiong (*Equal contribution)
🚀 **Code:** [AutoGEO on GitHub](https://github.com/cxcscmu/AutoGEO)
## Usage
This model is designed to be used through the [**AutoGEO framework**](https://github.com/cxcscmu/AutoGEO). Try it out in [huggingface Space](https://huggingface.co/spaces/cx-cmu/AutoGEO_Mini) or
Quick starts:
```python
from autogeo.rewriters import rewrite_document
rewritten_text = rewrite_document(
document="Input text.",
dataset="GEO-Bench",
engine_llm="gemini",
model_path="cx-cmu/AutoGEO_mini_Qwen1.7B_ResearchyGEO",
)
```
Evaluation:
```bash
python -m autogeo.evaluate \
--model autogeo_mini \
--model_path cx-cmu/AutoGEO_mini_Qwen1.7B_ResearchyGEO \
--dataset GEO-Bench
```
## Related Resources
* **Paper:** [https://arxiv.org/abs/2510.11438](https://arxiv.org/abs/2510.11438)
* **Code:** [https://github.com/cxcscmu/AutoGEO](https://github.com/cxcscmu/AutoGEO)
* **Dataset:** [https://huggingface.co/datasets/cx-cmu/GEO-Bench](https://huggingface.co/datasets/cx-cmu/GEO-Bench)
## Citation
If you use this model, please cite:
```bibtex
@article{wu2025generative,
title={What Generative Search Engines Like and How to Optimize Web Content Cooperatively},
author={Wu, Yujiang and Zhong, Shanshan and Kim, Yubin and Xiong, Chenyan},
journal={arXiv preprint arXiv:2510.11438},
year={2025}
}
``` |