[ICML 2025] SECOND: Mitigating Perceptual Hallucination in VLMs via Selective and Contrastive Decoding

This repository provides the official PyTorch implementation of the following paper:

SECOND: Mitigating Perceptual Hallucination in Vision-Language Models via Selective and Contrastive Decoding

📄 Paper (arXiv)
🌐 Project Page

---

🚀 Installation

Clone the repository and install dependencies:

cd SECOND/lmms-eval-vicuna
pip install -e .

cd LLaVA-NeXT
pip install -e .

---

🔧 Usage

Run SECOND with the provided scripts:

cd SECOND/lmms-eval-vicuna
./run.sh

---

📚 Citation

If you find this work useful, please consider citing:

@inproceedings{park2025second,
  title     = {SECOND: Mitigating Perceptual Hallucination in Vision-Language Models via Selective and Contrastive Decoding},
  author    = {Park, Woohyeon and Kim, Woojin and Kim, Jaeik and Do, Jaeyoung},
  booktitle = {Proceedings of the 42nd International Conference on Machine Learning (ICML)},
  year      = {2025},
  series    = {Proceedings of Machine Learning Research},
  publisher = {PMLR}
}