Pretrained Language Model Embryology: The Birth of ALBERT

Cheng-Han Chiang, Sung-Feng Huang, Hung-yi Lee


Abstract
While behaviors of pretrained language models (LMs) have been thoroughly examined, what happened during pretraining is rarely studied. We thus investigate the developmental process from a set of randomly initialized parameters to a totipotent language model, which we refer to as the embryology of a pretrained language model. Our results show that ALBERT learns to reconstruct and predict tokens of different parts of speech (POS) in different learning speeds during pretraining. We also find that linguistic knowledge and world knowledge do not generally improve as pretraining proceeds, nor do downstream tasks’ performance. These findings suggest that knowledge of a pretrained model varies during pretraining, and having more pretrain steps does not necessarily provide a model with more comprehensive knowledge. We provide source codes and pretrained models to reproduce our results at https://github.com/d223302/albert-embryology.
Anthology ID:
2020.emnlp-main.553
Volume:
Proceedings of the 2020 Conference on Empirical Methods in Natural Language Processing (EMNLP)
Month:
November
Year:
2020
Address:
Online
Editors:
Bonnie Webber, Trevor Cohn, Yulan He, Yang Liu
Venue:
EMNLP
SIG:
Publisher:
Association for Computational Linguistics
Note:
Pages:
6813–6828
Language:
URL:
https://aclanthology.org/2020.emnlp-main.553
DOI:
10.18653/v1/2020.emnlp-main.553
Bibkey:
Cite (ACL):
Cheng-Han Chiang, Sung-Feng Huang, and Hung-yi Lee. 2020. Pretrained Language Model Embryology: The Birth of ALBERT. In Proceedings of the 2020 Conference on Empirical Methods in Natural Language Processing (EMNLP), pages 6813–6828, Online. Association for Computational Linguistics.
Cite (Informal):
Pretrained Language Model Embryology: The Birth of ALBERT (Chiang et al., EMNLP 2020)
Copy Citation:
PDF:
https://aclanthology.org/2020.emnlp-main.553.pdf
Optional supplementary material:
 2020.emnlp-main.553.OptionalSupplementaryMaterial.zip
Video:
 https://slideslive.com/38938879
Code
 d223302/albert-embryology
Data
GLUEQNLI