arXiv：cs.LG· Louie Hong Yao, Vishesh Anand, Yuan Zhuang, Tianyu Jiang·· 5 小时前AI 评分29
LLM 表征中的反问句：一项线性探针研究
Rhetorical Questions in LLM Representations: A Linear Probing Study
AI 导读
研究用线性探针分析 LLM 对反问句的内部表征，发现反问信号在早期即出现，且由最后一 token 的表征最稳定地捕获。反问句与信息寻求问句在数据集内线性可分，跨数据集迁移时 AUROC 约 0.7-0.8，但不同数据集训练的探针在同一语料上排序重叠常低于 0.2。结果表明反问句由多个强调不同线索的线性方向编码，而非单一共享方向。
正文
Abstract:Rhetorical questions are asked not to seek information but to persuade or signal stance. How large language models internally represent them remains unclear. We analyze rhetorical questions in LLM representations using linear probes on two social-media datasets with different discourse contexts, and find that rhetorical signals emerge early and are most stably captured by last-token representations. Rhetorical questions are linearly separable from information-seeking questions within datasets, and remain detectable under cross-dataset transfer, reaching AUROC around 0.7-0.8. However, we demonstrate that transferability does not simply imply a shared representation. Probes trained on different datasets produce different rankings when applied to the same target corpus, with overlap among the top-ranked instances often below 0.2. Qualitative analysis shows that these divergences correspond to distinct rhetorical phenomena: some probes capture discourse-level rhetorical stance embedded in extended argumentation, while others emphasize localized, syntax-driven interrogative acts. Together, these findings suggest that rhetorical questions in LLM representations are encoded by multiple linear directions emphasizing different cues, rather than a single shared direction.
| Comments: | 18 pages, 15 figures, accepted to ACL 2026 |
| Subjects: | Computation and Language (cs.CL); Artificial Intelligence (cs.AI); Machine Learning (cs.LG) |
| Cite as: | arXiv:2604.14128 [cs.CL] |
| (or arXiv:2604.14128v3 [cs.CL] for this version) | |
| https://doi.org/10.48550/arXiv.2604.14128 arXiv-issued DOI via DataCite |
|
| Related DOI: | https://doi.org/10.18653/v1/2026.acl-long.5
DOI(s) linking to related resources |
Submission history
From: Hong Yao [view email]
[v1]
Wed, 15 Apr 2026 17:50:56 UTC (636 KB)
[v2]
Tue, 21 Apr 2026 23:20:36 UTC (636 KB)
[v3]
Fri, 2 Oct 2026 17:40:47 UTC (636 KB)
来源：arXiv：cs.LG · arxiv.org