Probing LLMs for Joint Encoding of Linguistic Categories

Giulio Starace, Konstantinos Papakostas, Rochelle Choenni, Apostolos Panagiotopoulos, Matteo Rosati, Alina Leidinger, Ekaterina Shutova

Submission Type: Regular Long Paper

Submission Track: Syntax, Parsing and their Applications

Submission Track 2: Interpretability, Interactivity, and Analysis of Models for NLP

Keywords: language model, transformer, interpretability, probing, syntax, pos, dependency, multilingual

TL;DR: We find evidence that LLMs jointly encode linguistic categories such as POS tags and dependency relations, across tasks and languages.

Abstract: 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.

Submission Number: 3436