LAFA: Agentic LLM-Driven Federated Analytics over Decentralized Data Sources

TL;DR

The paper tackles privacy concerns in LLM-based analytics by enabling natural-language queries over decentralized data without centralized data collection. It introduces LAFA, a hierarchical multi-agent framework that translates natural language into executable, privacy-preserving FA DAGs, using coarse-grained and fine-grained planners plus a DAG optimizer to minimize redundant operations. The approach leverages established FA backends (e.g., Arboretum, Orchard) and demonstrates near-perfect execution plan completion with substantial reductions in computation and communication overhead, validated on realistic query sets. The work advances practical privacy-preserving analytics by providing an end-to-end, NL-enabled compiler for FA workflows with quantified efficiency gains and robust planning behavior.

Abstract

Large Language Models (LLMs) have shown great promise in automating data analytics tasks by interpreting natural language queries and generating multi-operation execution plans. However, existing LLM-agent-based analytics frameworks operate under the assumption of centralized data access, offering little to no privacy protection. In contrast, federated analytics (FA) enables privacy-preserving computation across distributed data sources, but lacks support for natural language input and requires structured, machine-readable queries. In this work, we present LAFA, the first system that integrates LLM-agent-based data analytics with FA. LAFA introduces a hierarchical multi-agent architecture that accepts natural language queries and transforms them into optimized, executable FA workflows. A coarse-grained planner first decomposes complex queries into sub-queries, while a fine-grained planner maps each subquery into a Directed Acyclic Graph of FA operations using prior structural knowledge. To improve execution efficiency, an optimizer agent rewrites and merges multiple DAGs, eliminating redundant operations and minimizing computational and communicational overhead. Our experiments demonstrate that LAFA consistently outperforms baseline prompting strategies by achieving higher execution plan success rates and reducing resource-intensive FA operations by a substantial margin. This work establishes a practical foundation for privacy-preserving, LLM-driven analytics that supports natural language input in the FA setting.

Figures (3)

• Figure 1: The system overview of LAFA
• Figure 2: The workflow of LAFA using a query as an example
• Figure 3: The impact of DAG optimizer in reducing operation count