FinSight: Multi-Agent and Multi-Modal System for Equity Research and Reasoning
Abstract
Equity research is a significant business and primary focus of sell-side institutions such as investment banks. However, it has remained relatively untouched from enhancements and optimizations driven by LLMs based Multi-Agent systems. In this paper, we introduce FinSight - a multi-agent and multi-modal system for equity research and reasoning, which transforms sell-side research by bringing automation, consistency in quality and deeper research. Along with being the most comprehensive multi-agent system for equity research till date, it also makes significant contributions in Financial AI - introducing reasoning data and models for equity research for the first time and proposing a reasoning-model guided cross-analysis strategy which mimics a research analyst's cognitive abilities. Evaluations on standardized metrics establish its ability to perform accurate and comprehensive research, over existing systems.