跳到正文
研踪PaperTrail
← 研踪 PaperTrail

arXiv:2611.11239 · v1

A Layer-wise Study of How Style Rewriting Breaks the Calibration of Zero-Shot LLM Fake-News Detectors

首次提交 2026-11-08 16:48
版本更新 2026-11-08 16:48
雷达观测 2026-11-10 12:16
arXiv ↗PDF ↗

摘要

Large language models (LLMs) are increasingly asked to judge whether a news article is true, and their stated confidence is often taken at face value. We test what happens to that confidence when an article's writing style, but not its facts, is manipulated, as in recent \emph{style attacks} on fake-news detectors. Tracing how the prediction and its confidence form layer by layer inside the network, we study three model families (LLaMA-3-8B, Qwen2.5-7B, Mistral-7B) on three fake-news benchmarks, and find the same three patterns in every model--dataset pairing. First, on the original articles the models are competent and largely honest judges: they settle on an answer, briefly become overconfident, and then correct their confidence in the final layers. Second, style-rewritten articles break this self-correction: the models become far less accurate, under the strongest attack, worse than random guessing, yet remain just as confident, so they are \emph{confidently wrong} precisely when a warning is most needed. Third, nearly all of the damage comes from rewriting into sensational or emotionally charged styles, while sober, neutral rewrites leave the models' honesty largely intact. The practical implication is that an LLM's confidence about veracity cannot be trusted as a safety signal when an article's style may have been manipulated.

AI 论文解读

基于摘要 · 非全文审阅

此版本尚未生成 AI 解读。解读由后台任务生成并按论文版本保存,登录后共享已有结果。

版本差异

目前只收录了一个版本。收录下一版后,可以对照标题、作者、分类和摘要。

观测到的版本

相邻记录可能跳过中间版本;首次发现 v2 及以上也会标为版本更新。

v1 · 2026-11-08 16:48 · 最新观测版本

A Layer-wise Study of How Style Rewriting Breaks the Calibration of Zero-Shot LLM Fake-News Detectors

Large language models (LLMs) are increasingly asked to judge whether a news article is true, and their stated confidence is often taken at face value. We test what happens to that confidence when an article's writing style, but not its facts, is manipulated, as in recent \emph{style attacks} on fake-news detectors. Tracing how the prediction and its confidence form layer by layer inside the network, we study three model families (LLaMA-3-8B, Qwen2.5-7B, Mistral-7B) on three fake-news benchmarks, and find the same three patterns in every model--dataset pairing. First, on the original articles the models are competent and largely honest judges: they settle on an answer, briefly become overconfident, and then correct their confidence in the final layers. Second, style-rewritten articles break this self-correction: the models become far less accurate, under the strongest attack, worse than random guessing, yet remain just as confident, so they are \emph{confidently wrong} precisely when a warning is most needed. Third, nearly all of the damage comes from rewriting into sensational or emotionally charged styles, while sober, neutral rewrites leave the models' honesty largely intact. The practical implication is that an LLM's confidence about veracity cannot be trusted as a safety signal when an article's style may have been manipulated.

查看此版本解读 →查看此版本原文 ↗