Research map / Probabilistic and scientific modeling / Uncertainty and Gaussian processes
TabPFN: A Transformer That Solves Small Tabular Classification Problems in a Second
Read the paperCheck your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| Use an SCM prior instead of a BNN-only prior (ablation) | TabPFN · Prior ablation test set | Mean cross-entropy: 0.811±0.009 -> 0.771±0.006 | Helped |
| Use an SCM prior instead of a BNN-only prior (ablation) | TabPFN · Prior ablation test set | Mean ROC AUC: 0.865±0.007 -> 0.881±0.002 | Helped |
| Add a BNN prior to an SCM prior (ablation) | TabPFN · Prior ablation test set | Mean cross-entropy: 0.771±0.006 -> 0.776±0.009 | Hurt |
| Add a BNN prior to an SCM prior (ablation) | TabPFN · Prior ablation test set | Mean ROC AUC: 0.881±0.002 -> 0.883±0.003 | Helped |
| Add 32 data permutations for ensembling (ablation) | TabPFN · OpenML-CC18, 18 numerical datasets | Mean AUC OVO: 0.932±.0088 -> 0.934±.0086 | Helped |
| Add an AutoGluon prediction ensemble | TabPFN · OpenML-CC18, 30 datasets | Mean AUC OVO: 0.894±.01 -> 0.898±.0097 | Helped |
| Apply TabPFN to datasets with categorical features or missing values | TabPFN · OpenML-CC18 datasets | ROC AUC | Hurt |
| Add uninformative features | TabPFN · OpenML-CC18 numerical datasets | Classification accuracy | Hurt |
| Remove up to 30% of the least informative features | TabPFN · OpenML-CC18 numerical datasets | Classification accuracy | No clear effect |
| Rotate the feature space | TabPFN · OpenML-CC18 numerical datasets | ROC AUC | Hurt |
Closest papers by meaning
- TabPFN Unleashed: A Scalable and Effective Solution to Tabular Classification Problems (ICML 2025 · similarity 0.92)
- A Closer Look at TabPFN v2: Understanding Its Strengths and Extending Its Capabilities (NeurIPS 2025 · similarity 0.91)
- Drift-Resilient TabPFN: In-Context Learning Temporal Distribution Shifts on Tabular Data (NeurIPS 2024 · similarity 0.89)
- Retrieval & Fine-Tuning for In-Context Tabular Models (NeurIPS 2024 · similarity 0.89)
- TabICL: A Tabular Foundation Model for In-Context Learning on Large Data (ICML 2025 · similarity 0.88)
- TuneTables: Context Optimization for Scalable Prior-Data Fitted Networks (NeurIPS 2024 · similarity 0.88)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.