Exploring Internal Numeracy in Language Models: A Case Study on ALBERT

Ulme Wennberg, Gustav Eje Henter


Abstract
It has been found that Transformer-based language models have the ability to perform basic quantitative reasoning. In this paper, we propose a method for studying how these models internally represent numerical data, and use our proposal to analyze the ALBERT family of language models. Specifically, we extract the learned embeddings these models use to represent tokens that correspond to numbers and ordinals, and subject these embeddings to Principal Component Analysis (PCA). PCA results reveal that ALBERT models of different sizes, trained and initialized separately, consistently learn to use the axes of greatest variation to represent the approximate ordering of various numerical concepts. Numerals and their textual counterparts are represented in separate clusters, but increase along the same direction in 2D space. Our findings illustrate that language models, trained purely to model text, can intuit basic mathematical concepts, opening avenues for NLP applications that intersect with quantitative reasoning.
Anthology ID:
2024.mathnlp-1.5
Volume:
Proceedings of the 2nd Workshop on Mathematical Natural Language Processing @ LREC-COLING 2024
Month:
May
Year:
2024
Address:
Torino, Italia
Editors:
Marco Valentino, Deborah Ferreira, Mokanarangan Thayaparan, Andre Freitas
Venues:
MathNLP | WS
SIG:
Publisher:
ELRA and ICCL
Note:
Pages:
35–40
Language:
URL:
https://aclanthology.org/2024.mathnlp-1.5
DOI:
Bibkey:
Cite (ACL):
Ulme Wennberg and Gustav Eje Henter. 2024. Exploring Internal Numeracy in Language Models: A Case Study on ALBERT. In Proceedings of the 2nd Workshop on Mathematical Natural Language Processing @ LREC-COLING 2024, pages 35–40, Torino, Italia. ELRA and ICCL.
Cite (Informal):
Exploring Internal Numeracy in Language Models: A Case Study on ALBERT (Wennberg & Henter, MathNLP-WS 2024)
Copy Citation:
PDF:
https://aclanthology.org/2024.mathnlp-1.5.pdf