PluRel: Synthetic Data unlocks Scaling Laws for Relational Foundation Models

Relational Foundation Models (RFMs) facilitate data-driven decision-making by learning from complex multi-table databases. However, the diverse relational databases needed to train such models are rarely public due to privacy constraints. While there are methods to generate synthetic tabular data of arbitrary size, incorporating schema structure and primary–foreign key connectivity for multi-table generation remains challenging. Here we introduce PluRel, a framework to synthesize multi-tabular relational databases from scratch. In a step-by-step fashion, PluRel models (1) schemas with directed graphs, (2) inter-table primary-foreign key connectivity with bipartite graphs, and, (3) feature distributions in tables via conditional causal mechanisms. The design space across these stages supports the synthesis of a wide range of diverse databases, while being computationally lightweight. Using PluRel, we observe for the first time that (1) RFM pretraining loss exhibits power-law scaling with the number of synthetic databases and total pretraining tokens, (2) scaling the number of synthetic databases improves generalization to real databases, and (3) synthetic pretraining yields strong base models for continued pretraining on real databases. Overall, our framework and results position synthetic data scaling as a promising paradigm for RFMs.

💡 Research Summary

PluRel is a lightweight framework for synthesizing multi‑table relational databases from scratch, addressing the scarcity of public relational data needed to train Relational Foundation Models (RFMs). The authors decompose the generation process into three hierarchical stages. First, a database schema is sampled as a directed acyclic graph (DAG) where nodes represent tables and edges encode primary‑foreign key (P→F) relationships. Table metadata such as row counts and column counts are drawn independently, and tables are classified as “entity” (out‑degree ≥ 1) or “activity” (out‑degree = 0). Second, row‑level connectivity between parent and child tables is modeled using a Hierarchical Stochastic Block Model (HSBM). Rows are partitioned into hierarchical blocks; level‑wise probability matrices P