FISHNET: Financial Intelligence from Sub-querying, Harmonizing, Neural-Conditioning, Expert Swarms, and Task Planning

TL;DR

The invention of FISHNET (Financial Intelligence from Sub-querying, Harmonizing, Harmonizing, Neural-Conditioning, Expert swarming, and Task planning), an agentic architecture that accomplishes highly complex analytical tasks for more than 98,000 regulatory filings that vary immensely in terms of semantics, data hierarchy, or format.

Abstract

Financial intelligence generation from vast data sources has typically relied on traditional methods of knowledge-graph construction or database engineering. Recently, fine-tuned financial domain-specific Large Language Models (LLMs), have emerged. While these advancements are promising, limitations such as high inference costs, hallucinations, and the complexity of concurrently analyzing high-dimensional financial data, emerge. This motivates our invention FISHNET (Financial Intelligence from Sub-querying, Harmonizing, Neural-Conditioning, Expert swarming, and Task planning), an agentic architecture that accomplishes highly complex analytical tasks for more than 98,000 regulatory filings that vary immensely in terms of semantics, data hierarchy, or format. FISHNET shows remarkable performance for financial insight generation (61.8% success rate over 5.0% Routing, 45.6% RAG R-Precision). We conduct rigorous ablations to empirically prove the success of FISHNET, each agent's importance, and the optimized performance of assembling all agents. Our modular architecture can be leveraged for a myriad of use-cases, enabling scalability, flexibility, and data integrity that are critical for financial tasks.

Figures (2)

• Figure 1: Traditional Swarm Intelligence (SI) vs. FISHNET: SI relies on highly capable individual agents collectively working towards an efficient solution that exceeds the respective capabilities of those agents. SI typically excludes the presence of a central entity that harmonizes the agents' actions. Recently, a plethora of studies have explored an LLM's capability to orchestrate actions. In FISHNET, we combine these two approaches together; a central harmonizer can orchestrate while the expert agents also communicate.
• Figure 2: System Diagram of FISHNET: For any imposed query, (1) the query's hallucination probability is firstly calculated to improve on the query's quality. (2) Then, our sub-querying agent breaks down the query into different components. (3) Our task planning agent, leveraging basic In-Context Learning (ICL) and few-shot examples, devises the first iteration of the plan based on the sub-queries. (4) The plan is communicated to the expert agents and kicks off the swarming process - each expert agent starts its search. (5) The harmonizer synthesizes the information gathered via Swarm Intelligence (SI) and communicates it back to the task planning agent. (6) This enables the task planning agent to revise and optimize its previous plan. (7) The revised plan is then executed by the harmonizer agent and results are communicated back to the user.