APE: Faster and Longer Context-Augmented Generation via Adaptive Parallel Encoding

Installation

Install APE attention

python setup.py install

Install experiment dependencies

pip install -r requirements.txt

Demo

Here, we demonstrate a basic demo of APE in demo_APE.py

Experiments

To reproduce APE results on the retrieval-augmented generation (RAG) and in-context-learning (ICL) tasks, refer to the instruction and code available in the experiments directory.

Citation

@inproceedings{yang2025ape,
  title={APE: Faster and Longer Context-Augmented Generation via Adaptive Parallel Encoding},
  author={Yang, Xinyu and Chen, Tianqi and Chen, Beidi},
  booktitle={ICLR 2025},
  year={2025}
}