Gespeichert in:
Bibliographische Detailangaben
Hauptverfasser: Wang, Yuxin, Yu, Botao, Yang, Ivory, Hassanpour, Saeed, Vosoughi, Soroush
Format: Preprint
Veröffentlicht: 2025
Schlagworte:
Online-Zugang:https://arxiv.org/abs/2505.23914
Tags: Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
_version_ 1866910056726921216
author Wang, Yuxin
Yu, Botao
Yang, Ivory
Hassanpour, Saeed
Vosoughi, Soroush
author_facet Wang, Yuxin
Yu, Botao
Yang, Ivory
Hassanpour, Saeed
Vosoughi, Soroush
contents Large Language Models are widely used for content moderation but often present certain over-sensitivity, leading to misclassification of benign content and rejecting safe user commands. While previous research attributes this issue primarily to the presence of explicit offensive triggers, we statistically reveal a deeper connection beyond token level: When behaving over-sensitively, particularly on decontextualized statements, LLMs exhibit systematic topic-toxicity association patterns that go beyond explicit offensive triggers. To characterize these patterns, we propose Topic Association Analysis, a behavior-based probe that elicits short contextual scenarios for benign inputs and quantifies topic amplification between the scenario and the original comment. Across multiple LLMs and large-scale data, we find that more advanced models (e.g., GPT-4 Turbo) show stronger topic-association skew in false-positive cases despite lower overall false-positive rates. Moreover, via controlled prefix interventions, we show that topic cues can measurably shift false-positive rates, indicating that topic framing is decision-relevant. These results suggest that mitigating over-sensitivity may require addressing learned topic associations in addition to keyword-based filtering.
format Preprint
id arxiv_https___arxiv_org_abs_2505_23914
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle Probing Association Biases in LLM Moderation Over-Sensitivity
Wang, Yuxin
Yu, Botao
Yang, Ivory
Hassanpour, Saeed
Vosoughi, Soroush
Computation and Language
Artificial Intelligence
Large Language Models are widely used for content moderation but often present certain over-sensitivity, leading to misclassification of benign content and rejecting safe user commands. While previous research attributes this issue primarily to the presence of explicit offensive triggers, we statistically reveal a deeper connection beyond token level: When behaving over-sensitively, particularly on decontextualized statements, LLMs exhibit systematic topic-toxicity association patterns that go beyond explicit offensive triggers. To characterize these patterns, we propose Topic Association Analysis, a behavior-based probe that elicits short contextual scenarios for benign inputs and quantifies topic amplification between the scenario and the original comment. Across multiple LLMs and large-scale data, we find that more advanced models (e.g., GPT-4 Turbo) show stronger topic-association skew in false-positive cases despite lower overall false-positive rates. Moreover, via controlled prefix interventions, we show that topic cues can measurably shift false-positive rates, indicating that topic framing is decision-relevant. These results suggest that mitigating over-sensitivity may require addressing learned topic associations in addition to keyword-based filtering.
title Probing Association Biases in LLM Moderation Over-Sensitivity
topic Computation and Language
Artificial Intelligence
url https://arxiv.org/abs/2505.23914