LLM-based Multi-Agent System for Simulating Strategic and Goal-Oriented Data Marketplaces

TL;DR

The paper introduces LLM-MAS, a large language model based multi-agent system for simulating strategic and goal-oriented data marketplaces. By equipping buyer and seller agents with explicit objectives, natural language reasoning, and metadata-driven search via a vector database, the framework captures macro patterns such as trend emergence and scale-free transaction structures, while also offering micro-level insights into agent behaviors. Validation against Ocean Protocol data shows the model reproduces key structural and temporal market properties and demonstrates dynamics leading to domain-specific demand shifts. This approach advances market design study by bridging micro-agent decisions with macro market dynamics, enabling more realistic simulations and policy-relevant insights for data marketplaces.

Abstract

Data marketplaces, which mediate the purchase and exchange of data from third parties, have attracted growing attention for reducing the cost and effort of data collection while enabling the trading of diverse datasets. However, a systematic understanding of the interactions between market participants, data, and regulations remains limited. To address this gap, we propose a Large Language Model-based Multi-Agent System (LLM-MAS) for data marketplaces. In our framework, buyer and seller agents powered by LLMs operate with explicit objectives and autonomously perform strategic actions, such as planning, searching, purchasing, pricing, and updating data. These agents can reason about market dynamics, forecast future demand, and adjust strategies accordingly. Unlike conventional model-based simulations, which are typically constrained to predefined rules, LLM-MAS supports broader and more adaptive behavior selection through natural language reasoning. We evaluated the framework via simulation experiments using three distribution-based metrics: (1) the number of purchases per dataset, (2) the number of purchases per buyer, and (3) the number of repeated purchases of the same dataset. The results demonstrate that LLM-MAS more faithfully reproduces trading patterns observed in real data marketplaces compared to traditional approaches, and further captures the emergence and evolution of market trends.

Figures (12)

• Figure 1: Overall architecture of the proposed LLM-MAS for data marketplaces. The system consists of seller agents providing and updating data, and buyer agents pursuing goal-oriented acquisitions, a GoalGenerator for assigning analytical objectives, and a DataGenerator for producing metadata. Embedded metadata in a vector database supports cosine similarity-based search, reproducing the strategic behaviors, demand fluctuations, and trend emergence observed in real-world data marketplaces.
• Figure 2: Simulation process
• Figure 3: Three distributions of data transactions in a real-world data market and simulation results
• Figure 4: Number of data transactions, cumulative count of simulation components, and degree distribution of buyer-seller network.
• Figure 5: Heatmap for trend