Functional Interpolation for Relative Positions Improves Long Context Transformers
Abstract
Preventing the performance decay of Transformers on inputs longer than those used for training has been an important challenge in extending the context length of these models. Though the Transformer architecture has fundamentally no limits on the input sequence lengths it can process, the choice of position encoding used during training can limit the performance of these models on longer inputs. We propose a novel functional relative position encoding with progressive interpolation, FIRE, to improve Transformer generalization to longer contexts. We theoretically prove that this can represent some of the popular relative position encodings, such as T5's RPE, Alibi, and Kerple. We next empirically show that FIRE models have better generalization to longer contexts on both zero-shot language modeling and long text benchmarks.
Cite
Text
Li et al. "Functional Interpolation for Relative Positions Improves Long Context Transformers." International Conference on Learning Representations, 2024.Markdown
[Li et al. "Functional Interpolation for Relative Positions Improves Long Context Transformers." International Conference on Learning Representations, 2024.](https://mlanthology.org/iclr/2024/li2024iclr-functional/)BibTeX
@inproceedings{li2024iclr-functional,
title = {{Functional Interpolation for Relative Positions Improves Long Context Transformers}},
author = {Li, Shanda and You, Chong and Guruganesh, Guru and Ainslie, Joshua and Ontanon, Santiago and Zaheer, Manzil and Sanghai, Sumit and Yang, Yiming and Kumar, Sanjiv and Bhojanapalli, Srinadh},
booktitle = {International Conference on Learning Representations},
year = {2024},
url = {https://mlanthology.org/iclr/2024/li2024iclr-functional/}
}