通过概率曲线和动态语义水标识检测人工智能生成的文本
Yuhan Zhang1, Xingxiang Jiang2,3, Hua Sun2,3
1School of Computer Science and Artificial Intelligence, Nanjing University of Finance and Economics, Nanjing 210023, China.
Entropy (Basel, Switzerland)
|August 28, 2025
概括
检测人工智能生成的文字是很困难的. CurveMark使用概率曲率和水印来准确识别大型语言模型 (LLM) 内容,同时保持文本质量和抵御攻击.
科学领域:
- 人工智能
- 自然语言处理
- 信息安全
背景情况:
- 由于先进的AI文本生成,大型语言模型 (LLM) 面临重大内容认证挑战.
- 现有的检测方法难以获取有限的信息,
研究的目的:
- 开发一个强大的框架来检测来自LLM的AI生成文本.
- 克服当前检测方法的局限性,包括需要先前的模型知识.
主要方法:
- 推出了CurveMark,一个双通道检测框架,结合了概率曲率分析和动态语义水标.
- 使用信息理论原则来最大限度地提高文本来源和可观测特征之间的相互信息.
- 纳入贝叶斯的多假设检测框架,用于模型不可知的统计推理.
主要成果:
- 在多个数据集和LLM架构中实现95.4%的检测准确度.
- 证明质量降低最小,困惑度增加<1.3.
- 显示了对抗干扰的强大性能,保留了72-94%的信息.
结论:
- 对于人工智能生成的文本检测,CurveMark提供了一个高度准确和强大的解决方案.
- 该框架有效地平衡了检测准确性和文本质量以及对攻击的弹性.
- 这种方法在复杂的LLM时代提升了内容认证.
更多相关视频
相关概念视频
Non-equilibrium in the Cell
4.8K
An important concept in studying metabolism and energy is that of chemical equilibrium. Most chemical reactions are reversible. They can proceed in both directions, releasing energy into their environment in one direction, and absorbing it from the environment in the other direction. The same is true for the chemical reactions involved in cell metabolism, such as the breaking down and building up of proteins into and from individual amino acids, respectively. Reactants within a closed system...
4.8K
Detection of Black Holes
2.3K
Although black holes were theoretically postulated in the 1920s, they remained outside the domain of observational astronomy until the 1970s.
Their closest cousins are neutron stars, which are composed almost entirely of neutrons packed against each other, making them extremely dense. A neutron star has the same mass as the Sun but its diameter is only a few kilometers. Therefore, the escape velocity from their surface is close to the speed of light.
Not until the 1960s, when the first neutron...
Their closest cousins are neutron stars, which are composed almost entirely of neutrons packed against each other, making them extremely dense. A neutron star has the same mass as the Sun but its diameter is only a few kilometers. Therefore, the escape velocity from their surface is close to the speed of light.
Not until the 1960s, when the first neutron...
2.3K
Difference from Background: Limit of Detection
7.1K
The limit of detection (LOD) is the smallest amount of analyte that can be distinguished from the background noise. The LOD value corresponds to the concentration at which the analyte signal is three times larger than the standard deviation of the blank signal. Below this value, the analyte signal cannot be differentiated from the background noise. It is calculated by dividing the calibration slope by 3 times the standard deviation of the blank signals.
The LOD indicates the presence or absence...
The LOD indicates the presence or absence...
7.1K
Detection of Gross Error: The Q Test
6.4K
When one or more data points appear far from the rest of the data, there is a need to determine whether they are outliers and whether they should be eliminated from the data set to ensure an accurate representation of the measured value. In many cases, outliers arise from gross errors (or human errors) and do not accurately reflect the underlying phenomenon. In some cases, however, these apparent outliers reflect true phenomenological differences. In these cases, we can use statistical methods...
6.4K


