Memory-Augmented Agent Training for Business Document Understanding

TL;DR

This paper presents Matrix, a memory-augmented training paradigm for LLM agents that improves business document understanding by iterative learning and memory refinement, significantly outperforming baseline methods in transport reference extraction.

Contribution

The paper introduces Matrix, a novel memory-augmented training approach enabling LLM agents to develop domain expertise through experience-driven memory updates and iterative learning.

Findings

Matrix outperforms single LLM prompting by 30.3%.

Matrix outperforms vanilla LLM agent by 35.2%.

The agent system uses fewer API calls and can analyze longer documents.

Abstract

Traditional enterprises face significant challenges in processing business documents, where tasks like extracting transport references from invoices remain largely manual despite their crucial role in logistics operations. While Large Language Models offer potential automation, their direct application to specialized business domains often yields unsatisfactory results. We introduce Matrix (Memory-Augmented agent Training through Reasoning and Iterative eXploration), a novel paradigm that enables LLM agents to progressively build domain expertise through experience-driven memory refinement and iterative learning. To validate this approach, we collaborate with one of the world's largest logistics companies to create a dataset of Universal Business Language format invoice documents, focusing on the task of transport reference extraction. Experiments demonstrate that Matrix outperforms prompting a single LLM by 30.3%, vanilla LLM agent by 35.2%. We further analyze the metrics of the optimized systems and observe that the agent system requires less API calls, fewer costs and can analyze longer documents on average. Our methods establish a new approach to transform general-purpose LLMs into specialized business tools through systematic memory enhancement in document processing tasks.