README.md

---
license: apache-2.0
---

# VaPR-LLaVA-v1.5-13B
This model is an artifact of the work **VaPR – Vision-language Preference alignment for Reasoning (accepted at COLM 2025).**  📂 [Project Website](https://vap-r.github.io/)


## Base Model
- **Base Model**: [LLaVA v1.5 13B Instruct](https://huggingface.co/llava-hf/llava-1.5-7b-hf)

## Training Data
This model has been fine-tuned on the **VaPR-30k subset** dataset available at: [VaPR-30k](https://huggingface.co/datasets/VaPR-UCLA/vapr-30k)

## Citation

If you use this model, please cite the following paper:

```bibtex
inproceedings{
wadhawan2025vapr,
title={Va{PR} - Vision-language Preference alignment for Reasoning},
author={Rohan Wadhawan and Fabrice Y Harel-Canada and Zi-Yi Dou and Suhaila Shakiah and Robinson Piramuthu and Nanyun Peng},
booktitle={Second Conference on Language Modeling},
year={2025},
url={https://openreview.net/forum?id=uBAubFwymy}
}