随机鹦鹉寻找随机鹦鹉：LLMs 易调优且难以被其他 LLMs 检测出（英文版）

英文标题：Stochastic Parrots Looking for Stochastic Parrots: LLMs are Easy to Fine-Tune and Hard to Detect with other LLMs中文摘要：本文研究了如何对抗当前大规模语言模型检测工具的缺陷，发现攻击者结合 reinforcement from critic 优化方法和 AdamW 优化器可以轻松地规避检测，并对检测器进行破坏，这对防范恶意使用情况具有重要意义。英文摘要：The self-attention revolution allowed generative language models to scale andachieve increasingly impressive abilities. Such models - commonly referred toas Large Language Models (LLMs) - have recently gained prominence with thegeneral public, thanks to conver