---
license: other
task_categories:
- tabular-classification
- tabular-regression
language:
- en
tags:
- tabular
- synthetic
- real-data
- pretraining
- tabpfn
pretty_name: Tabula Pretraining Corpus
dataset_info:
  config_name: datagen_104
  features:
  - name: peer_rating
    dtype: float64
  - name: tenure_years
    dtype: float64
  - name: role_level
    dtype: float64
  - name: remote_days
    dtype: float64
  - name: team_size
    dtype: float64
  - name: absences
    dtype: float64
  - name: performance_score
    dtype: float64
  - name: promotions
    dtype: float64
  - name: overtime_hours
    dtype: float64
  - name: satisfaction
    dtype: float64
  - name: department_id
    dtype: float64
  - name: bonus_pct
    dtype: float64
  - name: salary
    dtype: float64
  - name: training_hours
    dtype: float64
  - name: attrition
    dtype: float64
  - name: hr_15
    dtype: float64
  - name: hr_16
    dtype: float64
  - name: hr_17
    dtype: float64
  - name: hr_18
    dtype: float64
  - name: target
    dtype: float64
  - name: _source_meta
    dtype: large_string
  - name: science_17
    dtype: float64
  - name: clicks
    dtype: float64
  - name: health_18
    dtype: float64
  - name: family_history
    dtype: float64
  - name: health_17
    dtype: float64
  - name: science_15
    dtype: float64
  - name: margin
    dtype: float64
  - name: fragile
    dtype: float64
  - name: finance_15
    dtype: float64
  - name: expenses
    dtype: float64
  - name: cost_usd
    dtype: float64
  - name: health_15
    dtype: float64
  - name: default
    dtype: float64
  - name: fever
    dtype: float64
  - name: volume
    dtype: float64
  - name: exercise_days
    dtype: float64
  - name: interest_rate
    dtype: float64
  - name: ecommerce_17
    dtype: float64
  - name: heart_rate
    dtype: float64
  - name: category
    dtype: float64
  - name: quantity
    dtype: float64
  - name: velocity
    dtype: float64
  - name: blood_pressure
    dtype: float64
  - name: damage_rate
    dtype: float64
  - name: temperature_k
    dtype: float64
  - name: employment_years
    dtype: float64
  - name: conversion
    dtype: float64
  - name: purity
    dtype: float64
  - name: price
    dtype: float64
  - name: reaction_time
    dtype: float64
  - name: credit_score
    dtype: float64
  - name: warehouse_id
    dtype: float64
  - name: route_id
    dtype: float64
  - name: reviews
    dtype: float64
  - name: smoking
    dtype: float64
  - name: load_factor
    dtype: float64
  - name: mass
    dtype: float64
  - name: weight_kg
    dtype: float64
  - name: pressure_pa
    dtype: float64
  - name: glucose
    dtype: float64
  - name: refund
    dtype: float64
  - name: carrier_id
    dtype: float64
  - name: science_16
    dtype: float64
  - name: cholesterol
    dtype: float64
  - name: priority
    dtype: float64
  - name: entropy
    dtype: float64
  - name: wavelength
    dtype: float64
  - name: ph
    dtype: float64
  - name: finance_17
    dtype: float64
  - name: fuel_consumption
    dtype: float64
  - name: liabilities
    dtype: float64
  - name: income
    dtype: float64
  - name: logistics_17
    dtype: float64
  - name: ecommerce_18
    dtype: float64
  - name: customs_flag
    dtype: float64
  - name: shipping_days
    dtype: float64
  - name: diagnosis
    dtype: float64
  - name: session_minutes
    dtype: float64
  - name: intensity
    dtype: float64
  - name: churn
    dtype: float64
  - name: finance_16
    dtype: float64
  - name: finance_18
    dtype: float64
  - name: hemoglobin
    dtype: float64
  - name: logistics_18
    dtype: float64
  - name: white_cell_count
    dtype: float64
  - name: assets
    dtype: float64
  - name: logistics_16
    dtype: float64
  - name: tax_rate
    dtype: float64
  - name: ecommerce_16
    dtype: float64
  - name: discount
    dtype: float64
  - name: health_16
    dtype: float64
  - name: age
    dtype: float64
  - name: distance_km
    dtype: float64
  - name: revenue
    dtype: float64
  - name: loan_amount
    dtype: float64
  - name: yield_pct
    dtype: float64
  - name: cart_size
    dtype: float64
  - name: balance
    dtype: float64
  - name: return_rate
    dtype: float64
  - name: debt_ratio
    dtype: float64
  - name: science_18
    dtype: float64
  - name: ecommerce_15
    dtype: float64
  - name: rating
    dtype: float64
  - name: alcohol_units
    dtype: float64
  - name: creatinine
    dtype: float64
  - name: concentration
    dtype: float64
  - name: logistics_15
    dtype: float64
  - name: acceleration
    dtype: float64
  - name: on_time
    dtype: float64
  - name: transit_days
    dtype: float64
  - name: return_flag
    dtype: float64
  - name: bmi
    dtype: float64
  - name: energy_kj
    dtype: float64
  splits:
  - name: train
    num_bytes: 21580119
    num_examples: 16500
  download_size: 2310682
  dataset_size: 21580119
configs:
- config_name: datagen_104
  data_files:
  - split: train
    path: datagen_104/train-*
---

# Tabula Pretraining Corpus

A continuously growing tabular pretraining corpus for the Tabula foundation model
(tabPFN-style in-context learning). Built by an autonomous agent that alternates
between harvesting permissively-licensed real datasets and generating high-quality
synthetic ones.

## Stats (auto-updated)

| Metric | Value |
|--------|-------|
| Total rows | 1,554,500 |
| Real-data batches | 0 |
| Synthetic batches | 42 |
| Last updated | 2026-03-12 15:17 UTC |

## Schema

Every row has feature columns plus `_source_meta` (JSON string):
- `batch_id`, `source_type`, `source_id`, `domain`, `task_type`, `license`, `citation_key`

## Sources & Citations

| batch_id | source_type | method | source_id | n_datasets | total_rows | status |
|----------|-------------|--------|-----------|------------|------------|--------|
| datagen_001 | synthetic | TreePrior | synthetic:TreePrior | 4 | 3500 | success |
| datagen_002 | synthetic | TreePrior | synthetic:TreePrior | 6 | 17000 | success |
| datagen_003 | synthetic | SCM | synthetic:SCM | 14 | 47000 | success |
| datagen_004 | synthetic | TreePrior | synthetic:TreePrior | 5 | 17000 | success |
| datagen_005 | synthetic | SCM | synthetic:SCM | 11 | 33000 | success |
| datagen_003 | synthetic | SCM | synthetic:SCM | 12 | 35500 | success |
| datagen_006 | synthetic | TreePrior | synthetic:TreePrior | 7 | 15500 | success |
| datagen_006 | synthetic | TreePrior | synthetic:TreePrior | 7 | 28000 | success |
| datagen_007 | synthetic | SCM | synthetic:SCM | 12 | 62500 | success |
| datagen_008 | synthetic | TreePrior | synthetic:TreePrior | 5 | 17500 | success |
| datagen_009 | synthetic | TreePrior | synthetic:TreePrior | 5 | 8500 | success |
| datagen_009 | synthetic | SCM | synthetic:SCM | 10 | 32000 | success |
| datagen_010 | synthetic | TreePrior | synthetic:TreePrior | 8 | 15500 | success |
| datagen_011 | synthetic | SCM | synthetic:SCM | 13 | 56500 | success |
| datagen_012 | synthetic | GaussianMixture | synthetic:GaussianMixture | 10 | 33500 | success |
| datagen_013 | synthetic | Polynomial | synthetic:Polynomial | 14 | 39000 | success |
| datagen_014 | synthetic | Regression | synthetic:Regression | 13 | 61500 | success |
| datagen_078 | synthetic | MixedType_TreePrior | synthetic:MixedType_TreePrior | 5 | 6500 | success |
| datagen_079 | synthetic | MixedType_SCM | synthetic:MixedType_SCM | 12 | 55000 | success |
| datagen_080 | synthetic | MixedType_GaussianMixture | synthetic:MixedType_GaussianMixture | 11 | 47000 | success |
| datagen_081 | synthetic | TreePrior | synthetic:TreePrior | 6 | 23500 | success |
| datagen_082 | synthetic | SCM | synthetic:SCM | 15 | 46000 | success |
| datagen_083 | synthetic | GaussianMixture | synthetic:GaussianMixture | 10 | 41000 | success |
| datagen_084 | synthetic | Polynomial | synthetic:Polynomial | 15 | 95500 | success |
| datagen_085 | synthetic | Regression | synthetic:Regression | 15 | 57500 | success |
| datagen_086 | synthetic | TimeSeries | synthetic:TimeSeries | 12 | 19500 | success |
| datagen_087 | synthetic | MixedType_TreePrior | synthetic:MixedType_TreePrior | 4 | 4000 | success |
| datagen_088 | synthetic | MixedType_SCM | synthetic:MixedType_SCM | 13 | 46500 | success |
| datagen_089 | synthetic | MixedType_GaussianMixture | synthetic:MixedType_GaussianMixture | 12 | 38000 | success |
| datagen_090 | synthetic | TreePrior | synthetic:TreePrior | 4 | 31000 | success |
| datagen_091 | synthetic | SCM | synthetic:SCM | 15 | 63000 | success |
| datagen_092 | synthetic | GaussianMixture | synthetic:GaussianMixture | 8 | 34000 | success |
| datagen_093 | synthetic | Polynomial | synthetic:Polynomial | 10 | 23000 | success |
| datagen_094 | synthetic | Regression | synthetic:Regression | 14 | 74000 | success |
| datagen_095 | synthetic | TimeSeries | synthetic:TimeSeries | 14 | 19500 | success |
| datagen_096 | synthetic | MixedType_TreePrior | synthetic:MixedType_TreePrior | 8 | 25000 | success |
| datagen_097 | synthetic | MixedType_SCM | synthetic:MixedType_SCM | 14 | 43500 | success |
| datagen_098 | synthetic | MixedType_GaussianMixture | synthetic:MixedType_GaussianMixture | 12 | 43000 | success |
| datagen_099 | synthetic | TreePrior | synthetic:TreePrior | 6 | 14500 | success |
| datagen_100 | synthetic | SCM | synthetic:SCM | 12 | 62500 | success |
| datagen_101 | synthetic | GaussianMixture | synthetic:GaussianMixture | 13 | 65500 | success |
| datagen_102 | synthetic | Polynomial | synthetic:Polynomial | 13 | 53000 | success |


## Key Citations

```bibtex
@inproceedings{hollmann2023tabpfn,
  title     = {TabPFN: A Transformer That Solves Small Tabular Classification Problems in a Second},
  author    = {Hollmann, Noah and M{\"u}ller, Samuel and Eggensperger, Katharina and Hutter, Frank},
  booktitle = {ICLR},
  year      = {2023}
}
@article{vanschoren2014openml,
  title   = {OpenML: Networked Science in Machine Learning},
  author  = {Vanschoren, Joaquin and van Rijn, Jan N. and Bischl, Bernd and Torgo, Luis},
  journal = {ACM SIGKDD Explorations},
  year    = {2014}
}
@article{scholkopf2021causal,
  title   = {Toward Causal Representation Learning},
  author  = {Sch{\"o}lkopf, Bernhard and others},
  journal = {Proceedings of the IEEE},
  year    = {2021}
}
```

## License

Individual rows carry their own `license` field inside `_source_meta`.
Synthetic rows are Apache 2.0. Real rows carry the original source license.