PluRel
star-project.stanford.edu
What PluRel is
PluRel is a free, open-source framework for synthesising relational and tabular data, aimed at relational foundation-model research and complex multi-table databases. It represents table schemas with directed graphs, inter-table links with bipartite graphs and feature distributions with conditional causal mechanisms. Configuration controls table layouts, table and row counts, column counts and structural causal-model parameters. An optional schema_file setting lets it generate data from an existing SQL schema. Generated columns support numeric, categorical and boolean values, with trends, cycles and fluctuations for temporal patterns. Dataset objects can be made compatible with RelBench. A multiprocessing script can generate multiple databases in parallel, with database and process counts configurable. Installation uses pip install plurel and requires Python 3.12 or later; the public repository lists an MIT licence. The project also states that synthetic pretraining alone is insufficient for robust zero-shot transfer, making continued real-data pretraining important for distribution alignment.
Who it suits
It suits researchers and data practitioners working with synthetic relational databases, temporal features and relational foundation models. It is intended for structured multi-table data, not unstructured data.
Good to know, and what to watch
- Open-source MIT licence
- Supports relational and tabular data
- Can use existing SQL schemas
- Models temporal feature patterns
- Parallel database generation available
- Requires Python 3.12 or later
- Synthetic pretraining alone lacks robust zero-shot transfer
- Does not target unstructured data
The verdict
PluRel provides configurable synthetic multi-table data generation with schema, relationship and temporal controls. It is a research-oriented framework, and its own stated transfer limitation means real-data continued pretraining remains important.
What the maker publishes
- Compatibility
- Generated datasets can be made compatible with RelBench. source
- Configurable output
- Its configuration controls table layouts, table and row counts, column counts, and structural causal model parameters. source
- Configurable output
- Its configuration controls table layouts, table and row counts, column counts, and structural causal model parameters. source
- Data types and patterns
- Feature generation supports numeric, categorical, and boolean values, with temporal trends, cycles, and fluctuations. source
- Feature types
- Generated feature columns support numeric, categorical, and boolean types. source
- Generation stages
- It models table schemas with directed graphs, inter-table connections with bipartite graphs, and feature distributions with conditional causal mechanisms. source
- Generation stages
- It models database schemas with directed graphs, inter-table primary–foreign key links with bipartite graphs, and feature distributions with conditional causal mechanisms. source
- Installation
- The library is installed with pip install plurel and requires Python 3.12 or later. source
- Installation requirement
- The library is installed with pip and requires Python 3.12 or later. source
- Intended context
- The project targets relational foundation model research and data-driven work with complex multi-table databases. source
- Intended use
- The project describes PluRel as a framework for research on relational foundation models and synthetic relational database generation. source
- Large-scale generation
- A multiprocessing script can generate databases in parallel, with the number of databases and processes configurable. source
- License
- The public GitHub repository lists an MIT license. source
- License
- The GitHub repository identifies the project license as MIT. source
- Parallel generation
- A multiprocessing-based script can generate databases in parallel, with the process count configurable. source
- Purpose
- PluRel is an open-source framework for synthesizing diverse relational and tabular data. source
- Purpose
- PluRel is an open-source library for synthesizing relational and tabular data. source
- RelBench compatibility
- The project says its generated dataset objects are compatible with RelBench. source
- Research results
- The project reports that scaling synthetic database diversity improves generalization to real databases and that synthetic pretraining can provide a base for continued pretraining on real data. source
- Research results
- The project reports that synthetic database diversity and pretraining token count both show power-law scaling in relational foundation model pretraining loss. source
- Schema input
- It can generate data from an existing SQL schema using the optional schema_file setting. source
- SQL schemas
- The configuration can optionally use an existing SQL schema as input. source
- Temporal patterns
- Feature generation can model temporal correlations using trend, cycle, and fluctuation components. source
- Transfer limitation
- The project states that synthetic pretraining alone is insufficient for robust zero-shot transfer and that continued pretraining on real data is critical for distribution alignment. source
Checked against the maker's pages on 8 Oct 2026.