LLMEmbed: Rethinking Lightweight LLM's Genuine Function in Text Classification

TL;DR

This paper introduces LLMEmbed, a simple transfer learning approach that leverages lightweight LLMs for efficient and robust text classification, achieving competitive performance with minimal resource consumption without fine-tuning.

Contribution

It proposes a novel lightweight embedding extraction and fusion strategy for LLMs, significantly reducing training overhead while maintaining high accuracy in text classification.

Findings

Achieves strong accuracy on benchmarks without fine-tuning.

Uses only 4% of model parameters and minimal resources compared to larger models.

Outperforms recent prompt-based methods in efficiency and effectiveness.

Abstract

With the booming of Large Language Models (LLMs), prompt-learning has become a promising method mainly researched in various research areas. Recently, many attempts based on prompt-learning have been made to improve the performance of text classification. However, most of these methods are based on heuristic Chain-of-Thought (CoT), and tend to be more complex but less efficient. In this paper, we rethink the LLM-based text classification methodology, propose a simple and effective transfer learning strategy, namely LLMEmbed, to address this classical but challenging task. To illustrate, we first study how to properly extract and fuse the text embeddings via various lightweight LLMs at different network depths to improve their robustness and discrimination, then adapt such embeddings to train the classifier. We perform extensive experiments on publicly available datasets, and the results show that LLMEmbed achieves strong performance while enjoys low training overhead using lightweight LLM backbones compared to recent methods based on larger LLMs, i.e. GPT-3, and sophisticated prompt-based strategies. Our LLMEmbed achieves adequate accuracy on publicly available benchmarks without any fine-tuning while merely use 4% model parameters, 1.8% electricity consumption and 1.5% runtime compared to its counterparts. Code is available at: https://github.com/ChunLiu-cs/LLMEmbed-ACL2024.