CLAILGOct 28, 2023

Probing LLMs for Joint Encoding of Linguistic Categories

arXiv:2310.18696v1135 citationsh-index: 13
Originality Incremental advance
AI Analysis

This provides insights into model interpretability for NLP researchers, but it is incremental as it builds on existing work on linguistic hierarchies in LLMs.

The paper tackles the problem of understanding how linguistic categories are jointly encoded in Large Language Models, finding evidence of shared representations for related part-of-speech classes and syntactic dependencies across languages.

Large Language Models (LLMs) exhibit impressive performance on a range of NLP tasks, due to the general-purpose linguistic knowledge acquired during pretraining. Existing model interpretability research (Tenney et al., 2019) suggests that a linguistic hierarchy emerges in the LLM layers, with lower layers better suited to solving syntactic tasks and higher layers employed for semantic processing. Yet, little is known about how encodings of different linguistic phenomena interact within the models and to what extent processing of linguistically-related categories relies on the same, shared model representations. In this paper, we propose a framework for testing the joint encoding of linguistic categories in LLMs. Focusing on syntax, we find evidence of joint encoding both at the same (related part-of-speech (POS) classes) and different (POS classes and related syntactic dependency relations) levels of linguistic hierarchy. Our cross-lingual experiments show that the same patterns hold across languages in multilingual LLMs.

Code Implementations1 repo
Foundations

The foundational work for this paper's niche, ranked by how specifically the neighbourhood builds on it — not by global fame.

Your Notes