FinSight: Towards Real-World Financial Deep Research

2025-10-19

AI Summary

Current AI systems struggle to autonomously generate professional-grade financial reports due to the need for integrated multi-source data acquisition, deep analytical reasoning, and high-fidelity multimodal presentation. To address this, we propose CAVM, a novel multi-agent framework featuring a variable memory mechanism that unifies data, tools, and agents within a shared programmable variable space. We introduce an iterative visual enhancement module to improve chart fidelity and a two-stage writing pipeline—comprising analytical chaining and reference-aware generation—to ensure logical coherence and contextual grounding. Key innovations include programmable variable-space orchestration, chain-of-analysis expansion, and citation-aware text generation. Evaluated on enterprise- and industry-level financial reporting tasks, CAVM achieves substantial improvements in factual accuracy, analytical depth, and visualization quality over state-of-the-art baselines, with overall report quality approaching human-expert standards.

Technology Category

• Cognitive Modeling & Cognitive Systems: Computational Creativity
• Computer Vision: Multi-modal Vision
• Machine Learning: Large Multimodal Models (LMMs)

Application Category

• Search and Retrieval-Augmented AI: Retrieval-Augmented Generation (RAG) and multi-modal RAG
• Semantics and Knowledge: Data modeling to support human-machine intelligence, including LLMs agents, intelligent system behavior, explanations, and user-friendly interactions
• Economics, Online Markets and Human Computation: Trust and reliance of crowd workers and data experts on GenAI

Abstract

Generating professional financial reports is a labor-intensive and intellectually demanding process that current AI systems struggle to fully automate. To address this challenge, we introduce FinSight (Financial InSight), a novel multi agent framework for producing high-quality, multimodal financial reports. The foundation of FinSight is the Code Agent with Variable Memory (CAVM) architecture, which unifies external data, designed tools, and agents into a programmable variable space, enabling flexible data collection, analysis and report generation through executable code. To ensure professional-grade visualization, we propose an Iterative Vision-Enhanced Mechanism that progressively refines raw visual outputs into polished financial charts. Furthermore, a two stage Writing Framework expands concise Chain-of-Analysis segments into coherent, citation-aware, and multimodal reports, ensuring both analytical depth and structural consistency. Experiments on various company and industry-level tasks demonstrate that FinSight significantly outperforms all baselines, including leading deep research systems in terms of factual accuracy, analytical depth, and presentation quality, demonstrating a clear path toward generating reports that approach human-expert quality.

Problem

• Automating labor-intensive financial report generation process
• Enhancing multimodal financial analysis through flexible agent architecture
• Improving visualization and writing quality in automated reports

Innovation

• CAVM architecture unifies data tools agents programmable space
• Iterative Vision-Enhanced Mechanism refines raw visual outputs charts
• Two stage Writing Framework expands analysis coherent multimodal reports