Evaluating LLM-Contaminated Crowdsourcing Data Without Ground Truth

Fuente: arXiv
Salvato in:
Dettagli Bibliografici
Autori principali: Zhang, Yichi, Pang, Jinlong, Zhu, Zhaowei, Liu, Yang
Natura: Preprint
Pubblicazione: 2025
Soggetti:
Accesso online:
Tags: Aggiungi Tag
Nessun Tag, puoi essere il primo ad aggiungerne!!
_version_ 1866912690920751104
author Zhang, Yichi
Pang, Jinlong
Zhu, Zhaowei
Liu, Yang
author_facet Zhang, Yichi
Pang, Jinlong
Zhu, Zhaowei
Liu, Yang
contents The recent success of generative AI highlights the crucial role of high-quality human feedback in building trustworthy AI systems. However, the increasing use of large language models (LLMs) by crowdsourcing workers poses a significant challenge: datasets intended to reflect human input may be compromised by LLM-generated responses. Existing LLM detection approaches often rely on high-dimensional training data such as text, making them unsuitable for annotation tasks like multiple-choice labeling. In this work, we investigate the potential of peer prediction -- a mechanism that evaluates the information within workers' responses without using ground truth -- to mitigate LLM-assisted cheating in crowdsourcing with a focus on annotation tasks. Our approach quantifies the correlations between worker answers while conditioning on (a subset of) LLM-generated labels available to the requester. Building on prior research, we propose a training-free scoring mechanism with theoretical guarantees under a crowdsourcing model that accounts for LLM collusion. We establish conditions under which our method is effective and empirically demonstrate its robustness in detecting low-effort cheating on real-world crowdsourcing datasets.
format Preprint
id arxiv_https___arxiv_org_abs_2506_06991
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle Evaluating LLM-Contaminated Crowdsourcing Data Without Ground Truth
Zhang, Yichi
Pang, Jinlong
Zhu, Zhaowei
Liu, Yang
Artificial Intelligence
Computer Science and Game Theory
Human-Computer Interaction
The recent success of generative AI highlights the crucial role of high-quality human feedback in building trustworthy AI systems. However, the increasing use of large language models (LLMs) by crowdsourcing workers poses a significant challenge: datasets intended to reflect human input may be compromised by LLM-generated responses. Existing LLM detection approaches often rely on high-dimensional training data such as text, making them unsuitable for annotation tasks like multiple-choice labeling. In this work, we investigate the potential of peer prediction -- a mechanism that evaluates the information within workers' responses without using ground truth -- to mitigate LLM-assisted cheating in crowdsourcing with a focus on annotation tasks. Our approach quantifies the correlations between worker answers while conditioning on (a subset of) LLM-generated labels available to the requester. Building on prior research, we propose a training-free scoring mechanism with theoretical guarantees under a crowdsourcing model that accounts for LLM collusion. We establish conditions under which our method is effective and empirically demonstrate its robustness in detecting low-effort cheating on real-world crowdsourcing datasets.
title Evaluating LLM-Contaminated Crowdsourcing Data Without Ground Truth
topic Artificial Intelligence
Computer Science and Game Theory
Human-Computer Interaction
url https://arxiv.org/abs/2506.06991