Hi! This is the repository for the EMNLP 2025 Findings paper:
Mixture of LoRA Experts for Continual Information Extraction with LLMs
Please make sure you have installed the packages by following instance:
conda env create -f environment.yaml
You can download the ED, RE, NER and replay data from here.
Not that we use the ACE and MAVEN datasets for ED evaluation. Due to that the ACE dataset are not released publicly, we can't provide the dataset after processing.
Download the LLaMA3.1-8b model and put it in ./pretrain_model
First obtain the training script.
python script/gen_script.pyThen run the resulting script to start the training and evaluation process.
sh MoLE-CIE.sh
We will release it soon. Thanks a lot!