TEXR: 36,115 Synthetic Tabular Datasets for Machine Learning Benchmarking

36,115 synthetic tabular datasets containing over 72 million rows were generated for the TEXR project. Each dataset contains 2,000 rows and between 5 and 33 features, paired with JSON metadata describing its topic, features, and Bayesian-network structure. The collection was created by eddyliu-hf and last updated on 2026-06-26.

Use Cases

• Benchmarking tabular data generation models based on the diverse topics and configurations described.
• Testing machine learning algorithms on synthetic data with known Bayesian-network structures.
• Studying the properties of synthetic datasets with varying feature schemas and value ranges.
• Developing methods for data augmentation using controlled synthetic data generation.

Strengths

• Large scale with 36,115 distinct datasets and over 72 million total rows.
• Structured metadata for each dataset includes topic, feature descriptions, value ranges, and Bayesian-network structure.
• Controlled generation with each dataset containing exactly 2,000 rows and between 5 and 33 features.

Limitations

• Column-level documentation is absent; field semantics must be inferred after download.
• Data is entirely synthetic, which may limit direct applicability to real-world problems.
• Last updated 2026-06-26 12:10:46; freshness should be verified.

Provenance

Source

huggingface

Collection Method

Synthetically generated for the TEXR project.

Freshness

2026-06-26 12:10:46