aiqbench analysis
This research addresses critical reliability gaps in tabular foundation models by introducing C-USIM, a method that ensures valid prediction intervals without requiring additional training or inference overhead. By leveraging split conformal prediction techniques, the study demonstrates how to mitigate undercoverage and overcoverage issues inherent in multimodal regression tasks using existing model outputs. The findings offer practical guidance for deploying trustworthy AI systems where uncertainty quantification is essential for decision-making.
Conformal Prediction and Conditional Coverage for Tabular Foundation Models
arXiv:2609.34887v1 Announce Type: cross 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.
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.
Machine Learning (stat.ML); Machine Learning (cs.LG)
arXiv:2609.34887 [stat.ML]
(or arXiv:2609.34887v1 [stat.ML] for this version)
https://doi.org/10.48550/arXiv.2609.34887 Focus to learn more arXiv-issued DOI via DataCite (pending registration)
arXivLabs is a framework that allows collaborators to develop and share new arXiv features directly on our website.
Both individuals and organizations that work with arXivLabs have embraced and accepted our values of openness, community, excellence, and user data privacy. arXiv is committed to these values and only works with partners that adhere to them.
Have an idea for a project that will add value for arXiv's community? Learn more about arXivLabs.
Full text under the publisher's open licence · 220 words
Related coverage
In-Context Residual Calibration for Uncertainty Quantification of Energy Time Series over Graphs
arXiv · Oct 8, 2026
Target leakage, not model class, explains reported accuracy in survey-based cardiovascular screening: a leakage-tiered audit of glass-box and tabular foundation models
arXiv · Sep 11, 2026
TabReliBench: Evaluating Tabular Foundation Models Beyond Predictive Accuracy
arXiv · Oct 12, 2026
Valid for Free: Homophily-Gated Conformal Prediction for Training-Free Node Classification with Tabular Foundation Models
arXiv · Oct 7, 2026