Identifying Linear Relational Concepts in Large Language Models

David Chanin, Anthony Hunter, Oana-Maria Camburu


Abstract
Transformer language models (LMs) have been shown to represent concepts as directions in the latent space of hidden activations. However, for any human-interpretable concept, how can we find its direction in the latent space? We present a technique called linear relational concepts (LRC) for finding concept directions corresponding to human-interpretable concepts by first modeling the relation between subject and object as a linear relational embedding (LRE). We find that inverting the LRE and using earlier object layers results in a powerful technique for finding concept directions that outperforms standard black-box probing classifiers. We evaluate LRCs on their performance as concept classifiers as well as their ability to causally change model output.
Anthology ID:
2024.naacl-long.85
Volume:
Proceedings of the 2024 Conference of the North American Chapter of the Association for Computational Linguistics: Human Language Technologies (Volume 1: Long Papers)
Month:
June
Year:
2024
Address:
Mexico City, Mexico
Editors:
Kevin Duh, Helena Gomez, Steven Bethard
Venue:
NAACL
SIG:
Publisher:
Association for Computational Linguistics
Note:
Pages:
1524–1535
Language:
URL:
https://aclanthology.org/2024.naacl-long.85
DOI:
Bibkey:
Cite (ACL):
David Chanin, Anthony Hunter, and Oana-Maria Camburu. 2024. Identifying Linear Relational Concepts in Large Language Models. In Proceedings of the 2024 Conference of the North American Chapter of the Association for Computational Linguistics: Human Language Technologies (Volume 1: Long Papers), pages 1524–1535, Mexico City, Mexico. Association for Computational Linguistics.
Cite (Informal):
Identifying Linear Relational Concepts in Large Language Models (Chanin et al., NAACL 2024)
Copy Citation:
PDF:
https://aclanthology.org/2024.naacl-long.85.pdf
Copyright:
 2024.naacl-long.85.copyright.pdf