Go back to publications View external

RiskLabs: Predicting Financial Risk Using Large Language Model Based on Multimodal and Multi-Sources Data

November 2025 Yupeng Cao, Zhi Chen, Qingyun Pei, Fabrizio Dimino, Lorenzo Ausiello, Prashant Kumar, K. P. Subbalakshmi, and Papa Momar Ndiaye ACM International Conference on AI in Finance (ICAIF 2024), Workshop on Multimodal Financial Foundation Models
#Large Language Models #Financial AI #Financial Risk Prediction #Multimodal Learning #Quantitative Finance

Abstract

The integration of Artificial Intelligence (AI) techniques, particularly Large Language Models (LLMs), into finance has attracted increasing attention in recent years. While previous studies have primarily focused on tasks such as financial text summarization, question answering, and stock movement prediction, financial risk prediction remains comparatively underexplored.

This paper introduces RiskLabs, a multimodal framework that leverages LLMs to analyze and predict financial risks by integrating heterogeneous financial information sources, including earnings conference calls, financial news, market-related time-series data, and contextual information surrounding corporate events. The framework combines multimodal representation learning, Retrieval-Augmented Generation (RAG), and deep learning techniques to forecast both market volatility and Value-at-Risk (VaR). Experimental results demonstrate that RiskLabs consistently outperforms conventional financial prediction models while providing richer financial representations through LLM-assisted reasoning.

Key Contributions

  1. Proposed a multimodal LLM framework for financial risk prediction by integrating market data, financial news, and earnings conference calls.
  2. Combined Retrieval-Augmented Generation (RAG), multimodal learning, and deep neural networks for financial reasoning and representation learning.
  3. Introduced a unified framework capable of predicting both market volatility and Value-at-Risk (VaR).
  4. Demonstrated that LLM-assisted multimodal financial analysis consistently improves predictive performance over conventional financial forecasting approaches.

Keywords

Large Language Models · Financial AI · Financial Risk Prediction · Multimodal Learning · Retrieval-Augmented Generation · Quantitative Finance