用稀疏自编码器构建可审计 AI 文本检测器
研究用稀疏自编码器将黑箱 AI 文本检测器拆解为可读、可审计的部分，在不削弱检测效果的同时，为每次判定给出可核验的加权写作线索。方法从微调 AI 文本检测器的 4 层激活值中提取特征，屏蔽缺乏可靠标签的特征，再由逻辑回归分类器完成判定。线索示例包括偏向 AI 的 em dashes，以及略微指向另一侧的“Firstly,”。
A black-box AI-text detector can be rebuilt from readable parts without losing its strength.
Each verdict then comes with weighted writing cues a reviewer can check, like em dashes pushing toward AI and “Firstly,” pulling slightly the other way.
activations from 4 layers of a fine-tuned AI-text detector are split into features by sparse autoencoders, features without reliable labels are masked out, and a logistic regression classifier makes the call from what remains.
– arxiv.org/abs/2610.30006
Title: "Building Auditable AI Text Detectors with Sparse Autoencoders"
来源：rohanpaul_ai · x.com