Pre-trained Large Language Models for Financial Sentiment Analysis

TL;DR

This paper investigates financial sentiment analysis for news titles and demonstrates that adapting the LLaMA-7B model with supervised fine-tuning yields state-of-the-art performance on the Financial PhraseBank benchmark, even without large-scale pretraining. It combines few-shot prompting, a dedicated SFT pipeline, and an optional classification-head approach to optimize sentiment classification among positive, negative, and neutral. Key findings show that SFT substantially improves accuracy (0.90) over a strong base, with ablation indicating that further pretraining offers limited gains. The work highlights the practical potential of task-adapted LLMs for finance-text analytics and points to scaling to larger models as future work.

Abstract

Financial sentiment analysis refers to classifying financial text contents into sentiment categories (e.g. positive, negative, and neutral). In this paper, we focus on the classification of financial news title, which is a challenging task due to a lack of large amount of training samples. To overcome this difficulty, we propose to adapt the pretrained large language models (LLMs) [1, 2, 3] to solve this problem. The LLMs, which are trained from huge amount of text corpora,have an advantage in text understanding and can be effectively adapted to domain-specific task while requiring very few amount of training samples. In particular, we adapt the open-source Llama2-7B model (2023) with the supervised fine-tuning (SFT) technique [4]. Experimental evaluation shows that even with the 7B model (which is relatively small for LLMs), our approach significantly outperforms the previous state-of-the-art algorithms.