Conformal Prediction and Conditional Coverage for Tabular Foundation Models
Abstract
Tabular foundation models (TFMs) provide predictive distributions for regression, but their prediction regions can exhibit undercoverage or overcoverage even when point predictions are accurate. We introduce C-USIM (Conditionally-Uniformized Score Integration Method), a lightweight application of highest predictive density split conformal prediction that accommodates multimodal predictions. Given calibration and test outputs, it requires no additional training or model inference. It provides finite-sample marginal validity under our assumptions. We bound conditional-marginal coverage gaps using distribution-estimation error and score discreteness, and examine coverage heterogeneity through percentile rank-score plots. Experiments with TabPFN and TabICL show improved marginal coverage accuracy and lower average conditional and group coverage errors. Under a fixed data budget, allocating more observations to calibration can reduce marginal coverage error despite less accurate point predictions.
Related
- How Synthetic Labels Improve Conformal Prediction: A Perspective on Conditional Coverage
- Sharp training-conditional coverage for conformal prediction under covariate shift
- Benchmarking Tabular Foundation Models for Conditional Density Estimation in Regression
- SurvFM enables tabular foundation models for right-censored survival prediction
- Tabular foundation models are not a shortcut to calibrated clinical risk prediction: a pre-registered simulation against the true risk
BibTeX
@misc{park2026conformal,
title = {{Conformal Prediction and Conditional Coverage for Tabular Foundation Models}},
author = {Sungwoo Park and Sunghee Park and Won Chang},
howpublished = {arXiv},
year = {2026},
month = sep,
doi = {10.48550/arxiv.2609.34887},
eprint = {2609.34887},
archivePrefix = {arXiv},
url = {https://arxiv.org/abs/2609.34887},
note = {CC BY 4.0},
}
RIS
TY - UNPB
TI - Conformal Prediction and Conditional Coverage for Tabular Foundation Models
AU - Sungwoo Park
AU - Sunghee Park
AU - Won Chang
PY - 2026
DA - 2026/09/28
PB - arXiv
DO - 10.48550/arxiv.2609.34887
UR - https://arxiv.org/abs/2609.34887
AB - Tabular foundation models (TFMs) provide predictive distributions for regression, but their prediction regions can exhibit undercoverage or overcoverage even when point predictions are accurate. We introduce C-USIM (Conditionally-Uniformized Score Integration Method), a lightweight application of highest predictive density split conformal prediction that accommodates multimodal predictions. Given calibration and test outputs, it requires no additional training or model inference. It provides finite-sample marginal validity under our assumptions. We bound conditional-marginal coverage gaps using distribution-estimation error and score discreteness, and examine coverage heterogeneity through percentile rank-score plots. Experiments with TabPFN and TabICL show improved marginal coverage accuracy and lower average conditional and group coverage errors. Under a fixed data budget, allocating more observations to calibration can reduce marginal coverage error despite less accurate point predictions.
ER -