跳到正文
HuggingFace Daily Papers(社區熱門論文)·· 2026-10-21AI 評分40

利用稀疏自編碼器構建可審計的 AI 文字偵測器

Building Auditable AI Text Detectors with Sparse Autoencoders

AI 導讀

研究團隊提出一套利用稀疏自編碼器(SAE)的流程,從神經網絡偵測器中提取可解讀的語言特徵,以人類可讀的概念分類 AI 生成文字。研究發現,基於這些特徵的分類器具備較強可解讀性,對分佈外文本的表現亦優於神經網絡及文體特徵分類器。

正文

Published on Oct 21

Authors:

,

,

,

,

Abstract

A pipeline using sparse autoencoders extracts interpretable linguistic features from neural detectors, yielding more accurate and explainable AI-generated text classification.

AI-generated text detectors are increasingly deployed in consequential settings---peer review, education, admissions, journalism, and content moderation---where their outputs directly affect individuals. Yet the strongest detectors are fine-tuned neural classifiers whose decisions are opaque, while transparent alternatives (stylometric and linguistic-feature classifiers) underperform them by wide margins. We present a pipeline that converts a fine-tuned neural detector into a classifier that operates on human-readable concepts. We accomplish this by training sparse autoencoders (SAEs) directly on a detector's residual stream, assigning auto-interpretation labels to each SAE feature, and retaining only features whose generated descriptions pass a global fidelity threshold. We then train a lightweight classifier on just the interpretable features. We find that our SAEs recover latent concepts that correspond to highly specific and predictive linguistic "tells" of AI-generated text. Classifiers built on these extracted features are not only highly interpretable but exhibit stronger performance on out-of-distribution texts than either neural or stylometric alternatives.

View arXiv page View PDF GitHub 10 Add to collection

Get this paper in your agent:

hf papers read 2610.30006

Don't have the latest CLI?

curl -LsSf https://hf.co/cli/install.sh | bash

Models citing this paper 3

liamdugan/detector-scope

Text Classification •

0.3B •

Updated 4 days ago

•

318

liamdugan/detector-scope-peer-review-gpt4o

Text Classification •

0.3B •

Updated 4 days ago

•

36

liamdugan/detector-scope-peer-review-llama

Text Classification •

0.3B •

Updated 4 days ago

•

22

Datasets citing this paper 0

No dataset linking this paper

Cite arxiv.org/abs/2610.30006 in a dataset README.md to link it from this page.

Spaces citing this paper 1

Collections including this paper 1

來源:HuggingFace Daily Papers(社區熱門論文) · huggingface.co