Update README.md
Browse files
README.md
CHANGED
|
@@ -1,3 +1,59 @@
|
|
| 1 |
---
|
| 2 |
license: apache-2.0
|
|
|
|
|
|
|
|
|
|
|
|
|
| 3 |
---
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
---
|
| 2 |
license: apache-2.0
|
| 3 |
+
datasets:
|
| 4 |
+
- iLearn-Lab/NeurIPS25-SymMPO
|
| 5 |
+
base_model:
|
| 6 |
+
- liuhaotian/llava-v1.5-7b
|
| 7 |
---
|
| 8 |
+
|
| 9 |
+
|
| 10 |
+
<div align="center">
|
| 11 |
+
<h2 align="center"> <b>SymMPO: Mitigating Hallucination Through Theory-Consistent Symmetric Multimodal Preference Optimization</b>
|
| 12 |
+
</h2>
|
| 13 |
+
|
| 14 |
+
|
| 15 |
+
<a target="_blank" href="https://github.com/Liuwq-bit">Wenqi Liu</a><sup>1</sup>,
|
| 16 |
+
<a target="_blank" href="https://scholar.google.com/citations?user=29gP4okAAAAJ">Xuemeng Song</a><sup>2</sup>,
|
| 17 |
+
<a target="_blank" href="https://scholar.google.com/citations?user=tvPOeFQAAAAJ">Jiaxi Li</a><sup>3</sup>,
|
| 18 |
+
<a target="_blank" href="https://scholar.google.com/citations?user=im-bS2YAAAAJ">Yinwei Wei</a><sup>1</sup>,
|
| 19 |
+
<a target="_blank" href="https://scholar.google.com/citations?user=VWunnXEAAAAJ">Zheng Na</a><sup>4</sup>,
|
| 20 |
+
<a target="_blank" href="https://scholar.google.com/citations?user=aZZfn90AAAAJ">Jianhua Yin</a><sup>1</sup>,
|
| 21 |
+
<a target="_blank" href="https://scholar.google.com/citations?user=yywVMhUAAAAJ">Liqiang Nie</a><sup>5</sup>
|
| 22 |
+
<br>
|
| 23 |
+
<sup>1</sup>Shandong University   
|
| 24 |
+
<sup>2</sup>Southern University of Science and Technology   
|
| 25 |
+
<sup>3</sup>University of Georgia   
|
| 26 |
+
<br>
|
| 27 |
+
<sup>4</sup>National University of Singapore   
|
| 28 |
+
<sup>5</sup>Harbin Institute of Technology, Shenzhen   
|
| 29 |
+
<br />
|
| 30 |
+
|
| 31 |
+
<a href='https://arxiv.org/abs/2506.11712'><img src='https://img.shields.io/badge/Paper-Arxiv-red'></a>
|
| 32 |
+
<a href='https://huggingface.co/iLearn-Lab/NeurIPS25-SymMPO-7B'><img src='https://img.shields.io/badge/Model-7B-yellow'></a>
|
| 33 |
+
<a href='https://huggingface.co/iLearn-Lab/NeurIPS25-SymMPO-13B'><img src='https://img.shields.io/badge/Model-13B-yellow'></a>
|
| 34 |
+
<a href='https://huggingface.co/datasets/iLearn-Lab/NeurIPS25-SymMPO'><img src='https://img.shields.io/badge/Dataset-HF-blue'></a>
|
| 35 |
+
|
| 36 |
+
</div>
|
| 37 |
+
|
| 38 |
+
|
| 39 |
+
---
|
| 40 |
+
|
| 41 |
+
## Introduction
|
| 42 |
+
|
| 43 |
+
We present **SymMPO**, a framework for mitigating hallucination in multimodal large language models (MLLMs). Our method introduces a theory-consistent symmetric multimodal preference optimization approach that addresses the hallucination problem from a principled perspective. This repository provides the official implementation, pretrained checkpoints, and evaluation scripts built on top of [LLaVA](https://github.com/haotian-liu/LLaVA).
|
| 44 |
+
|
| 45 |
+
|
| 46 |
+
## Citation
|
| 47 |
+
|
| 48 |
+
If you find our work helpful, please consider citing:
|
| 49 |
+
|
| 50 |
+
```bibtex
|
| 51 |
+
@inproceedings{
|
| 52 |
+
liu2025mitigating,
|
| 53 |
+
title={Mitigating Hallucination Through Theory-Consistent Symmetric Multimodal Preference Optimization},
|
| 54 |
+
author={Wenqi Liu and Xuemeng Song and Jiaxi Li and Yinwei Wei and Na Zheng and Jianhua Yin and Liqiang Nie},
|
| 55 |
+
booktitle={The Thirty-ninth Annual Conference on Neural Information Processing Systems},
|
| 56 |
+
year={2025},
|
| 57 |
+
url={https://openreview.net/forum?id=tIW29IpCwG}
|
| 58 |
+
}
|
| 59 |
+
```
|