You can deploy this model and start earning money today!

iLearn-Lab/
NeurIPS25-SymMPO-7B

Name: NeurIPS25-SymMPO-7B
Rating: 2.5 (11 reviews)

safetensors

llava_llama

iLearn-Lab/NeurIPS25-SymMPO

2506.11712

liuhaotian/llava-v1.5-7b

finetune

apache-2.0

Views

No views yet

SymMPO: Mitigating Hallucination Through Theory-Consistent Symmetric Multimodal Preference Optimization

Wenqi Liu¹, Xuemeng Song², Jiaxi Li³, Yinwei Wei¹, Zheng Na⁴, Jianhua Yin¹, Liqiang Nie⁵
¹Shandong University ²Southern University of Science and Technology ³University of Georgia
⁴National University of Singapore ⁵Harbin Institute of Technology, Shenzhen

Introduction

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.

Citation

If you find our work helpful, please consider citing:

bibtex

@inproceedings{
  liu2025mitigating,
  title={Mitigating Hallucination Through Theory-Consistent Symmetric Multimodal Preference Optimization},
  author={Wenqi Liu and Xuemeng Song and Jiaxi Li and Yinwei Wei and Na Zheng and Jianhua Yin and Liqiang Nie},
  booktitle={The Thirty-ninth Annual Conference on Neural Information Processing Systems},
  year={2025},
  url={https://openreview.net/forum?id=tIW29IpCwG}
}

iLearn-Lab/NeurIPS25-SymMPO-7B

SymMPO: Mitigating Hallucination Through Theory-Consistent Symmetric Multimodal Preference Optimization

Introduction

Citation

iLearn-Lab/
NeurIPS25-SymMPO-7B