This work has been accepted as the long paper ''LH-Mix: Local Hierarchy Correlation Guided Mixup over Hierarchical Prompt Tuning'' in KDD 2025.
The code is based on "HPT: Hierarchy-aware Prompt Tuning for Hierarchical Text Classification". We appreciate their sharing.
Please refer to https://github.com/wzh9969/HPT.
You can start LH-Mix directly by running the following code:
python train.py --name public --batch 16 --data rcv1 --threshold 0.75 --alpha 1 --warm_up 0
@article{kong2024lhmix,
title={LH-Mix: Local Hierarchy Correlation Guided Mixup over Hierarchical Prompt Tuning},
author={Fanshuang Kong, Richong Zhang and Ziqiao Wang},
booktitle={Proceedings of the ACM SIGKDD Conference on Knowledge Discovery and Data Mining},
year={2025},
}