What do LLMs Know about Financial Markets? A Case Study on Reddit Market Sentiment Analysis

TL;DR

This paper demonstrates how large language models can generate weak labels for Reddit market sentiment analysis, enabling the training of a smaller, production-ready model that performs comparably to supervised models without extensive labeled data.

Contribution

The study introduces a semi-supervised approach using LLMs to generate financial sentiment labels, improving label stability and model performance with minimal prompting.

Findings

Prompting LLMs with Chain-of-Thought improves label accuracy.

Forcing multiple reasoning paths enhances label stability.

The distilled model matches supervised model performance.

Abstract

Market sentiment analysis on social media content requires knowledge of both financial markets and social media jargon, which makes it a challenging task for human raters. The resulting lack of high-quality labeled data stands in the way of conventional supervised learning methods. Instead, we approach this problem using semi-supervised learning with a large language model (LLM). Our pipeline generates weak financial sentiment labels for Reddit posts with an LLM and then uses that data to train a small model that can be served in production. We find that prompting the LLM to produce Chain-of-Thought summaries and forcing it through several reasoning paths helps generate more stable and accurate labels, while using a regression loss further improves distillation quality. With only a handful of prompts, the final model performs on par with existing supervised models. Though production applications of our model are limited by ethical considerations, the model's competitive performance points to the great potential of using LLMs for tasks that otherwise require skill-intensive annotation.