Characterizing Verbatim Short-Term Memory in Neural Language Models

Kristijan Armeni, Christopher Honey, Tal Linzen


Abstract
When a language model is trained to predict natural language sequences, its prediction at each moment depends on a representation of prior context. What kind of information about the prior context can language models retrieve? We tested whether language models could retrieve the exact words that occurred previously in a text. In our paradigm, language models (transformers and an LSTM) processed English text in which a list of nouns occurred twice. We operationalized retrieval as the reduction in surprisal from the first to the second list. We found that the transformers retrieved both the identity and ordering of nouns from the first list. Further, the transformers’ retrieval was markedly enhanced when they were trained on a larger corpus and with greater model depth. Lastly, their ability to index prior tokens was dependent on learned attention patterns. In contrast, the LSTM exhibited less precise retrieval, which was limited to list-initial tokens and to short intervening texts. The LSTM’s retrieval was not sensitive to the order of nouns and it improved when the list was semantically coherent. We conclude that transformers implemented something akin to a working memory system that could flexibly retrieve individual token representations across arbitrary delays; conversely, the LSTM maintained a coarser and more rapidly-decaying semantic gist of prior tokens, weighted toward the earliest items.
Anthology ID:
2022.conll-1.28
Volume:
Proceedings of the 26th Conference on Computational Natural Language Learning (CoNLL)
Month:
December
Year:
2022
Address:
Abu Dhabi, United Arab Emirates (Hybrid)
Venue:
CoNLL
SIG:
Publisher:
Association for Computational Linguistics
Note:
Pages:
405–424
Language:
URL:
https://aclanthology.org/2022.conll-1.28
DOI:
Bibkey:
Cite (ACL):
Kristijan Armeni, Christopher Honey, and Tal Linzen. 2022. Characterizing Verbatim Short-Term Memory in Neural Language Models. In Proceedings of the 26th Conference on Computational Natural Language Learning (CoNLL), pages 405–424, Abu Dhabi, United Arab Emirates (Hybrid). Association for Computational Linguistics.
Cite (Informal):
Characterizing Verbatim Short-Term Memory in Neural Language Models (Armeni et al., CoNLL 2022)
Copy Citation:
PDF:
https://preview.aclanthology.org/author-url/2022.conll-1.28.pdf