ProbabilityBench: Model Evaluation and Calibration with Exact Ground-Truth Uncertainty
Abstract
Evaluating model uncertainty on real-world data is challenging because each instance provides only a single observed label, leaving true conditional class probabilities unknown. To address this, we introduce ProbabilityBench, a benchmark that provides exact, instance-level class probabilities across three simulated text-based domains: peer review, covert information transfer, and credit risk. By explicitly defining the full data-generating process for each environment, we analytically compute both exact class probabilities and Bayes error. Furthermore, we provide a theoretical guarantee that rendering these cases into natural-language memos strictly preserves their underlying generative probabilities. This exactness enables direct measurement of reported probability quality via KL divergence. Across 16 evaluated models, most models on the observed cost-quality Pareto frontiers are decision models, including Jev and Perplexity Decider. Finally, ProbabilityBench serves as a meta-evaluation framework, leveraging known ground-truth values to benchmark the accuracy of standard calibration metrics and Bayes-error estimators themselves.
Related
- A Three-Way Testing Framework for Quantifying Epistemic Calibration Uncertainty in SBI
- Calibration Assessment for Nested Uncertainty Sets
- Prediction Uncertainty Controlled Computer Model Calibration
- Dynamic Mission Risk Management via Adaptive Information Acquisition and Online Uncertainty Calibration
- Federated Calibration of Motion Uncertainty for UAV Tracking in Multi-Operator ISAC Systems
BibTeX
@misc{dai2026probabilitybench,
title = {{ProbabilityBench: Model Evaluation and Calibration with Exact Ground-Truth Uncertainty}},
author = {Yuyang Dai and Yuting Tang and Futoshi Futami and Takashi Ishida},
howpublished = {arXiv},
year = {2026},
month = oct,
doi = {10.48550/arxiv.2610.17016},
eprint = {2610.17016},
archivePrefix = {arXiv},
url = {https://arxiv.org/abs/2610.17016},
note = {CC BY 4.0},
}
RIS
TY - UNPB
TI - ProbabilityBench: Model Evaluation and Calibration with Exact Ground-Truth Uncertainty
AU - Yuyang Dai
AU - Yuting Tang
AU - Futoshi Futami
AU - Takashi Ishida
PY - 2026
DA - 2026/10/12
PB - arXiv
DO - 10.48550/arxiv.2610.17016
UR - https://arxiv.org/abs/2610.17016
AB - Evaluating model uncertainty on real-world data is challenging because each instance provides only a single observed label, leaving true conditional class probabilities unknown. To address this, we introduce ProbabilityBench, a benchmark that provides exact, instance-level class probabilities across three simulated text-based domains: peer review, covert information transfer, and credit risk. By explicitly defining the full data-generating process for each environment, we analytically compute both exact class probabilities and Bayes error. Furthermore, we provide a theoretical guarantee that rendering these cases into natural-language memos strictly preserves their underlying generative probabilities. This exactness enables direct measurement of reported probability quality via KL divergence. Across 16 evaluated models, most models on the observed cost-quality Pareto frontiers are decision models, including Jev and Perplexity Decider. Finally, ProbabilityBench serves as a meta-evaluation framework, leveraging known ground-truth values to benchmark the accuracy of standard calibration metrics and Bayes-error estimators themselves.
ER -