Towards Robust Low-Resource Fine-Tuning with Multi-View Compressed Representations

Linlin Liu, Xingxuan Li, Megh Thakkar, Xin Li, Shafiq Joty, Luo Si, Lidong Bing


Abstract
Due to the huge amount of parameters, finetuning of pretrained language models (PLMs) is prone to overfitting in the low resource scenarios. In this work, we present a novel method that operates on the hidden representations of a PLM to reduce overfitting. During fine-tuning, our method inserts random autoencoders between the hidden layers of a PLM, which transform activations from the previous layers into multi-view compressed representations before feeding them into the upper layers. The autoencoders are plugged out after fine-tuning, so our method does not add extra parameters or increase computation cost during inference. Our method demonstrates promising performance improvement across a wide range of sequence- and token-level lowresource NLP tasks.
Anthology ID:
2023.acl-long.264
Volume:
Proceedings of the 61st Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers)
Month:
July
Year:
2023
Address:
Toronto, Canada
Editors:
Anna Rogers, Jordan Boyd-Graber, Naoaki Okazaki
Venue:
ACL
SIG:
Publisher:
Association for Computational Linguistics
Note:
Pages:
4799–4816
Language:
URL:
https://aclanthology.org/2023.acl-long.264
DOI:
10.18653/v1/2023.acl-long.264
Bibkey:
Cite (ACL):
Linlin Liu, Xingxuan Li, Megh Thakkar, Xin Li, Shafiq Joty, Luo Si, and Lidong Bing. 2023. Towards Robust Low-Resource Fine-Tuning with Multi-View Compressed Representations. In Proceedings of the 61st Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers), pages 4799–4816, Toronto, Canada. Association for Computational Linguistics.
Cite (Informal):
Towards Robust Low-Resource Fine-Tuning with Multi-View Compressed Representations (Liu et al., ACL 2023)
Copy Citation:
PDF:
https://aclanthology.org/2023.acl-long.264.pdf
Video:
 https://aclanthology.org/2023.acl-long.264.mp4