HLPD: 通过人类语言偏好对齐大语言模型以检测机器修订文本
HLPD: Aligning LLMs to Human Language Preference for Machine-Revised Text Detection
AI总结：
HLPD通过人类语言偏好优化提升机器修订文本检测性能，实现对GPT系列模型和先进大语言模型生成文本的高效识别。
摘要：
为防止由大语言模型生成的可信内容所引发的虚假信息和社会问题，开发高效且可靠的方法来识别文本来源至关重要。先前的方法在检测完全由大语言模型生成的文本方面表现出色。然而，这些方法在面对更高级的大语言模型输出或具有对抗性多任务机器修订的文本时表现不佳，尤其是在黑盒设置中，生成模型是未知的。为了解决这一挑战，基于人类写作具有独特风格模式的假设，我们提出了人类语言偏好检测（HLPD）。HLPD采用基于奖励的对齐过程，即人类语言偏好优化（HLPO），将评分模型的标记分布转向人类样式的写作，使模型更敏感于人类写作，从而增强对机器修订文本的识别能力。我们在对抗性多任务评估框架中测试了HLPD，该框架利用五维提示生成器和多个先进的大语言模型来创建多样的修订场景。当检测由GPT系列模型修订的文本时，HLPD在AUROC上比ImBD提高了15.11%，超过Fast-DetectGPT 45.56%。当评估由先进大语言模型生成的文本时，HLPD实现了最高的平均AUROC，比ImBD高5.53%，比Fast-DetectGPT高34.14%。代码将在https://github.com/dfq2021/HLPD上公开。
英文摘要：
To prevent misinformation and social issues arising from trustworthy-looking content generated by LLMs, it is crucial to develop efficient and reliable methods for identifying the source of texts. Previous approaches have demonstrated exceptional performance in detecting texts fully generated by LLMs. However, these methods struggle when confronting more advanced LLM output or text with adversarial multi-task machine revision, especially in the black-box setting, where the generating model is unknown. To address this challenge, grounded in the hypothesis that human writing possesses distinctive stylistic patterns, we propose Human Language Preference Detection (HLPD). HLPD employs a reward-based alignment process, Human Language Preference Optimization (HLPO), to shift the scoring model's token distribution toward human-like writing, making the model more sensitive to human writing, therefore enhancing the identification of machine-revised text. We test HLPD in an adversarial multi-task evaluation framework that leverages a five-dimensional prompt generator and multiple advanced LLMs to create diverse revision scenarios. When detecting texts revised by GPT-series models, HLPD achieves a 15.11% relative improvement in AUROC over ImBD, surpassing Fast-DetectGPT by 45.56%. When evaluated on texts generated by advanced LLMs, HLPD achieves the highest average AUROC, exceeding ImBD by 5.53% and Fast-DetectGPT by 34.14%. Code will be made available at https://github.com/dfq2021/HLPD.
机构、总结、中文摘要均由AI分析而来，可能存在错误，请以论文正文为准。
