Revisiting Word Embeddings in the LLM Era

TL;DR

This paper systematically compares classical and LLM-induced embeddings, revealing that LLMs excel in semantic clustering and analogy tasks, but classical models often outperform in sentence similarity assessments, questioning the assumed superiority of LLM embeddings.

Contribution

It provides a comprehensive comparison between classical and LLM-induced embeddings, clarifying their relative strengths and limitations in various NLP tasks.

Findings

LLMs cluster semantically related words more tightly.

LLMs perform better on analogy tasks in decontextualized settings.

Classical models like SimCSE outperform LLMs in sentence similarity tasks.

Abstract

Large Language Models (LLMs) have recently shown remarkable advancement in various NLP tasks. As such, a popular trend has emerged lately where NLP researchers extract word/sentence/document embeddings from these large decoder-only models and use them for various inference tasks with promising results. However, it is still unclear whether the performance improvement of LLM-induced embeddings is merely because of scale or whether underlying embeddings they produce significantly differ from classical encoding models like Word2Vec, GloVe, Sentence-BERT (SBERT) or Universal Sentence Encoder (USE). This is the central question we investigate in the paper by systematically comparing classical decontextualized and contextualized word embeddings with the same for LLM-induced embeddings. Our results show that LLMs cluster semantically related words more tightly and perform better on analogy tasks in decontextualized settings. However, in contextualized settings, classical models like SimCSE often outperform LLMs in sentence-level similarity assessment tasks, highlighting their continued relevance for fine-grained semantics.