Customized FinGPT Search Agents Using Foundation Models

TL;DR

This paper introduces customized FinGPT search agents tailored for individual and institutional users, leveraging retrieval-augmented generation and fine-tuning to improve accuracy, relevance, and response time in financial data analysis.

Contribution

It presents novel methods for customizing FinGPT agents for different user types using RAG and fine-tuning, addressing privacy and timeliness in financial applications.

Findings

Outperform existing models in accuracy and relevance

Achieve faster response times in financial data retrieval

Effectively handle privacy and data sensitivity issues

Abstract

Current large language models (LLMs) have proven useful for analyzing financial data, but most existing models, such as BloombergGPT and FinGPT, lack customization for specific user needs. In this paper, we address this gap by developing FinGPT Search Agents tailored for two types of users: individuals and institutions. For individuals, we leverage Retrieval-Augmented Generation (RAG) to integrate local documents and user-specified data sources. For institutions, we employ dynamic vector databases and fine-tune models on proprietary data. There are several key issues to address, including data privacy, the time-sensitive nature of financial information, and the need for fast responses. Experiments show that FinGPT agents outperform existing models in accuracy, relevance, and response time, making them practical for real-world applications.