Matrix: Peer-to-Peer Multi-Agent Synthetic Data Generation Framework

TL;DR

Matrix is a decentralized, peer-to-peer framework for multi-agent synthetic data generation that scales efficiently and improves throughput without sacrificing quality.

Contribution

It introduces a scalable, flexible, and decentralized multi-agent synthesis framework built on Ray, eliminating the need for a central orchestrator.

Findings

Matrix achieves 2-15x higher data throughput across various scenarios.

The framework scales to tens of thousands of concurrent workflows.

Matrix maintains output quality while increasing efficiency.

Abstract

Synthetic data has become increasingly important for training large language models, especially when real data is scarce, expensive, or privacy-sensitive. Many such generation tasks require coordinated multi-agent workflows, where specialized agents collaborate to produce data that is higher quality, more diverse, and structurally richer. However, existing frameworks for multi-agent synthesis often depend on a centralized orchestrator, creating scalability bottlenecks, or are hardcoded for specific domains, limiting flexibility. We present \textbf{Matrix}, a decentralized framework that represents both control and data flow as serialized messages passed through distributed queues. This peer-to-peer design eliminates the central orchestrator. Each task progresses independently through lightweight agents, while compute-intensive operations, such as LLM inference or containerized environments, are handled by distributed services. Built on Ray, Matrix scales to tens of thousands of concurrent agentic workflows and provides a modular, configurable design that enables easy adaptation to a wide range of data generation workflows. We evaluate Matrix across diverse synthesis scenarios, such as multi-agent collaborative dialogue, web-based reasoning data extraction, and tool-use trajectory generation in customer service environments. In all cases, Matrix achieves $2$--$15\times$ higher data generation throughput under identical hardware resources, without compromising output quality.