References & Citations
Computer Science > Computation and Language
Title: Detecting Conceptual Abstraction in LLMs
(Submitted on 24 Apr 2024 (v1), last revised 25 Apr 2024 (this version, v2))
Abstract: We present a novel approach to detecting noun abstraction within a large language model (LLM). Starting from a psychologically motivated set of noun pairs in taxonomic relationships, we instantiate surface patterns indicating hypernymy and analyze the attention matrices produced by BERT. We compare the results to two sets of counterfactuals and show that we can detect hypernymy in the abstraction mechanism, which cannot solely be related to the distributional similarity of noun pairs. Our findings are a first step towards the explainability of conceptual abstraction in LLMs.
Submission history
From: Alhassan Abdelhalim [view email][v1] Wed, 24 Apr 2024 12:52:45 GMT (1659kb,D)
[v2] Thu, 25 Apr 2024 23:11:20 GMT (1653kb,D)
Link back to: arXiv, form interface, contact.