Concordia: Self-Improving Synthetic Tables for Federated LLMs
2026-05-11 • Machine Learning
Machine Learning
AI summaryⓘ
The authors address the challenge of training large language models on private tabular data across different clients without sharing any raw data. They propose Concordia, a method that uses synthetic data generated and improved locally to help models learn better while respecting privacy. Their approach lets clients adapt models efficiently and adjust the synthetic data based on private feedback, coordinating improvements across clients without sharing sensitive data. Experiments on real-world financial and healthcare data show that this method works better and is more stable than previous approaches that treat synthetic data generation and model training separately.
Federated LearningLarge Language ModelsSynthetic DataNon-IID DataParameter-Efficient Fine-TuningLoRAPolicy OptimizationTabular DataPrivacy PreservationDistribution Shift
Authors
Jimin Huang, Duanyu Feng, Nuo Chen, Xiaoyu Wang, Zhiqiang Zhang, Xueqing Peng, Mingquan Lin, Prayag Tiwari, Guojun Xiong, Alejandro Lopez-Lira, Sophia Ananiadou
Abstract
Federated learning (FL) enables training large language models (LLMs) without sharing raw data, but adapting LLMs under strict data isolation and non-IID client distributions remains challenging in practice. Synthetic data offers a natural privacy-preserving surrogate for local training, yet existing federated pipelines typically treat synthetic generation as static or loosely coupled with downstream optimization, leading to rapidly diminishing utility under heterogeneous clients. We study federated adaptation of LLMs on tabular tasks where raw records and validation data cannot be shared, and local training must rely entirely on synthetic tables. We propose Concordia, a tri-level optimization framework that aligns synthetic data generation with federated validation utility despite these constraints. At the client level, models are adapted via parameter-efficient LoRA training on synthetic tables. Clients additionally learn lightweight utility scorers from private validation feedback to reweight synthetic samples during local training. At the outer level, each client refines its own synthetic table generator using group-relative policy optimization (GRPO), guided by an ensemble of heterogeneous scorers shared across clients, without aggregating generator parameters or exposing validation data. Experiments on privacy-sensitive tabular benchmarks from finance and healthcare demonstrate that Concordia consistently improves federated performance, cross-client stability, and robustness to distribution shift compared to static and decoupled synthetic-data baselines.