Two Conformal Constructions for Adaptive Within-Document AI-Text Screening
Abstract
We study false-alert control when screening for text generated by artificial intelligence (AI). The screening procedure selects document prefixes and detectors from observed evidence and may stop before exhausting its inspection budget. We give two finite-sample constructions under document-level exchangeability between human calibration documents and a new null document, with no restriction on dependence among tokens within a document. Construction A registers a finite family of prefix-detector scores and allocates a false-alert budget across their conformal ranks. A union bound protects any executed subset of that family. Construction B calibrates the complete-path maximum of a development-fixed adaptive policy. Each partial-path maximum is bounded by the complete maximum, so a terminal conformal rank protects early stopping without splitting the error budget. We prove marginal control of any false alert across the permitted inspection path and derive necessary calibration counts for rejection. We also state oracle testing, distribution-shift, and independent-audit bounds with their additional assumptions. Both constructions protect stopping within their specified scope; neither proof constructs an e-process or justifies multiplying conformal ranks. Detection power and computational savings remain questions for empirical evaluation.
Related
- Change detection with conformal martingales: new optimal constructions, and suboptimality of existing methods
- Finite-sample certification of coverage regret for budgeted document selection under adaptive stratified auditing
- Adaptive Conformal Inference through the Lens of Blackwell Approachability
- SPACR: Single-Pass Adaptive Training of Uncertainty-Aware Conformal Regressors
- Adaptive conformal prediction on the Grassmannian
BibTeX
@misc{mandap2026conformal,
title = {{Two Conformal Constructions for Adaptive Within-Document AI-Text Screening}},
author = {Marco Mandap and Jerahmeel Hipolito and Arcel Galvez and Charlie Margaret Balagtas and Michael Joshua Buluran and Jeff Roel Durmiendo and Rizzette E. Lopez},
howpublished = {arXiv},
year = {2026},
month = sep,
doi = {10.48550/arxiv.2609.31547},
eprint = {2609.31547},
archivePrefix = {arXiv},
url = {https://arxiv.org/abs/2609.31547},
note = {CC BY 4.0},
}
RIS
TY - UNPB
TI - Two Conformal Constructions for Adaptive Within-Document AI-Text Screening
AU - Marco Mandap
AU - Jerahmeel Hipolito
AU - Arcel Galvez
AU - Charlie Margaret Balagtas
AU - Michael Joshua Buluran
AU - Jeff Roel Durmiendo
AU - Rizzette E. Lopez
PY - 2026
DA - 2026/09/25
PB - arXiv
DO - 10.48550/arxiv.2609.31547
UR - https://arxiv.org/abs/2609.31547
AB - We study false-alert control when screening for text generated by artificial intelligence (AI). The screening procedure selects document prefixes and detectors from observed evidence and may stop before exhausting its inspection budget. We give two finite-sample constructions under document-level exchangeability between human calibration documents and a new null document, with no restriction on dependence among tokens within a document. Construction A registers a finite family of prefix-detector scores and allocates a false-alert budget across their conformal ranks. A union bound protects any executed subset of that family. Construction B calibrates the complete-path maximum of a development-fixed adaptive policy. Each partial-path maximum is bounded by the complete maximum, so a terminal conformal rank protects early stopping without splitting the error budget. We prove marginal control of any false alert across the permitted inspection path and derive necessary calibration counts for rejection. We also state oracle testing, distribution-shift, and independent-audit bounds with their additional assumptions. Both constructions protect stopping within their specified scope; neither proof constructs an e-process or justifies multiplying conformal ranks. Detection power and computational savings remain questions for empirical evaluation.
ER -