arXiv cs.CL· SungHo Kim, SangKeun Lee·· 12 小时前AI 评分18
KoLing：面向韩语 LLM 生成文本检测的偏见控制对抗基准与 KoLingProbe 检测器
Beyond Surface Cues: Robust LLM-Generated Korean Text Detection
AI 导读
研究团队提出 KoLing，一个包含 50.9k 实例的偏见控制对抗基准，用于抑制表层线索并分别评估 LLM 生成与 LLM 修改的韩语文本检测。同时提出基于理论语言学、建模韩语多层级特性的可解释检测器 KoLingProbe，在所有检测任务上显著优于强基线且对对抗攻击保持稳健。分析显示，融合多层级互补语言学特征可同时提升性能与可解释性。
正文 · 原文
Abstract:As large language models (LLMs) generate increasingly fluent text, the need for reliable detection of LLM-generated content has grown substantially. However, existing research remains heavily English-centric, leaving Korean largely underexplored. Moreover, current evaluations of Korean text detection often rely on superficial cues, leading to overly optimistic estimates of detection performance. To address these limitations, we introduce \textsc{KoLing}, a bias-controlled adversarial benchmark comprising 50.9k instances that suppresses superficial shortcuts while preserving semantic content and separately evaluates LLM-generated and LLM-modified text detection. We further propose KoLingProbe, an interpretable detector grounded in theoretical linguistics that models Korean-specific properties across multiple linguistic levels. Experiments show that KoLingProbe substantially outperforms strong baselines on all detection tasks while remaining robust to adversarial attacks. Our analyses further reveal that integrating complementary linguistic features across multiple levels yields both improved performance and interpretability.
| Subjects: | Computation and Language (cs.CL) |
| Cite as: | arXiv:2610.22940 [cs.CL] |
| (or arXiv:2610.22940v1 [cs.CL] for this version) | |
| https://doi.org/10.48550/arXiv.2610.22940 arXiv-issued DOI via DataCite (pending registration) |
Submission history
From: SungHo Kim [view email]
[v1]
Fri, 16 Oct 2026 05:07:15 UTC (325 KB)
来源：arXiv cs.CL · arxiv.org