Calibration and Risk-Budgeted Decisions with System One Models
Abstract
LLM workflows and agents rely on a set of binary decisions like allow/block an input, or discard/retain a retrieved document. When human deferral is unavailable, uncertain cases must still receive an action. We study System One decision scores in this setting by defaulting to a conservative fallback when uncertain, and controlling the false omission risk within the subset of inputs assigned the utility-maximizing action. Across six datasets covering moderation, safety, and reranking, we compare general System One models with task baselines and a local open-weights fine-tuned System One model. We show that naive baselines for threshold selection exceed the target risk in roughly half of settings with defined test risk, compared with 7.8\% for Learn Then Test (LTT), which leverages finite-sample statistical guarantees. Our results show why relying on a probability-like interface is insufficient for deployment decisions and how risk-budgeted selection trades useful action against observed error.
Related
- Hierarchical Utility Calibration for Structured Multiclass Decisions
- Sequential Certification of Threshold Decisions in Rare-Event Risk Prediction
- The optimal operation decisions of expert service enterprises based on queuing system
- Recertifying decisions via e-closure: uniform improvements of conformal risk control and randomized e-BH
- Does Calibration Pick the Right Model? Likelihood Evidence from Income Risk
BibTeX
@misc{messoudi2026calibration,
title = {{Calibration and Risk-Budgeted Decisions with System One Models}},
author = {Soundouss Messoudi and Youness Mansar},
howpublished = {arXiv},
year = {2026},
month = oct,
doi = {10.48550/arxiv.2610.19811},
eprint = {2610.19811},
archivePrefix = {arXiv},
url = {https://arxiv.org/abs/2610.19811},
}
RIS
TY - UNPB
TI - Calibration and Risk-Budgeted Decisions with System One Models
AU - Soundouss Messoudi
AU - Youness Mansar
PY - 2026
DA - 2026/10/14
PB - arXiv
DO - 10.48550/arxiv.2610.19811
UR - https://arxiv.org/abs/2610.19811
AB - LLM workflows and agents rely on a set of binary decisions like allow/block an input, or discard/retain a retrieved document. When human deferral is unavailable, uncertain cases must still receive an action. We study System One decision scores in this setting by defaulting to a conservative fallback when uncertain, and controlling the false omission risk within the subset of inputs assigned the utility-maximizing action. Across six datasets covering moderation, safety, and reranking, we compare general System One models with task baselines and a local open-weights fine-tuned System One model. We show that naive baselines for threshold selection exceed the target risk in roughly half of settings with defined test risk, compared with 7.8\% for Learn Then Test (LTT), which leverages finite-sample statistical guarantees. Our results show why relying on a probability-like interface is insufficient for deployment decisions and how risk-budgeted selection trades useful action against observed error.
ER -