Latent Preference Coding

---

Setup

Env

git clone https://github.com/AntResearchNLP/LPC.git
cd LPC
pip install -r requirements.txt

Data

Please download UltraFeedback from huggingface. Then modify DATA_PATH in scripts/train_lpc.sh.

Training

sh scripts/train_lpc.sh 

Evaluation

We use Language Model Evaluation Harness for model evaluation.

@article{gong2025latent,
  title={Latent preference coding: Aligning large language models via discrete latent codes},
  author={Gong, Zhuocheng and Guan, Jian and Wu, Wei and Zhang, Huishuai and Zhao, Dongyan},
  journal={arXiv preprint arXiv:2505.04993},
  year={2025}
}