How Sampling Affects the Detectability of Machine-written texts: A Comprehensive Study

Fuente: arXiv
Saved in:
Bibliographic Details
Main Authors: Dubois, Matthieu, Yvon, François, Piantanida, Pablo
Format: Preprint
Published: 2025
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866908595506905088
author Dubois, Matthieu
Yvon, François
Piantanida, Pablo
author_facet Dubois, Matthieu
Yvon, François
Piantanida, Pablo
contents As texts generated by Large Language Models (LLMs) are ever more common and often indistinguishable from human-written content, research on automatic text detection has attracted growing attention. Many recent detectors report near-perfect accuracy, often boasting AUROC scores above 99\%. However, these claims typically assume fixed generation settings, leaving open the question of how robust such systems are to changes in decoding strategies. In this work, we systematically examine how sampling-based decoding impacts detectability, with a focus on how subtle variations in a model's (sub)word-level distribution affect detection performance. We find that even minor adjustments to decoding parameters - such as temperature, top-p, or nucleus sampling - can severely impair detector accuracy, with AUROC dropping from near-perfect levels to 1\% in some settings. Our findings expose critical blind spots in current detection methods and emphasize the need for more comprehensive evaluation protocols. To facilitate future research, we release a large-scale dataset encompassing 37 decoding configurations, along with our code and evaluation framework https://github.com/BaggerOfWords/Sampling-and-Detection
format Preprint
id arxiv_https___arxiv_org_abs_2510_13681
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle How Sampling Affects the Detectability of Machine-written texts: A Comprehensive Study
Dubois, Matthieu
Yvon, François
Piantanida, Pablo
Computation and Language
As texts generated by Large Language Models (LLMs) are ever more common and often indistinguishable from human-written content, research on automatic text detection has attracted growing attention. Many recent detectors report near-perfect accuracy, often boasting AUROC scores above 99\%. However, these claims typically assume fixed generation settings, leaving open the question of how robust such systems are to changes in decoding strategies. In this work, we systematically examine how sampling-based decoding impacts detectability, with a focus on how subtle variations in a model's (sub)word-level distribution affect detection performance. We find that even minor adjustments to decoding parameters - such as temperature, top-p, or nucleus sampling - can severely impair detector accuracy, with AUROC dropping from near-perfect levels to 1\% in some settings. Our findings expose critical blind spots in current detection methods and emphasize the need for more comprehensive evaluation protocols. To facilitate future research, we release a large-scale dataset encompassing 37 decoding configurations, along with our code and evaluation framework https://github.com/BaggerOfWords/Sampling-and-Detection
title How Sampling Affects the Detectability of Machine-written texts: A Comprehensive Study
topic Computation and Language
url https://arxiv.org/abs/2510.13681