arXiv cs.CL· Nhi Hoai Doan, Tatsuya Hiraoka, Munachiso Samuel Nwadike, Kentaro Inui·· 23 小时前AI 评分22
极性组合假说:来自 NLI 否定现象的因果证据
Polarity Composition Hypothesis: Causal Evidence from Negation in NLI
AI 导读
研究提出“极性组合假说”这一高层因果模型,解释语言模型在自然语言推理(NLI)中如何处理否定,行为评测显示模型预测大体符合该假说。团队用 Distributed Alignment Search 找到与假说变量对齐、具有因果效力的子空间,这些表征可跨否定形式迁移,并部分迁移至 MNLI 子集。数据集与代码已公开,论文被 EACL 2027 主会接收。
正文 · 原文
Abstract:We investigate how language models process negation in natural language inference (NLI). We propose the \textit{Polarity Composition Hypothesis}, a high-level causal model of this computation. Behavioral evaluations show that language-model predictions largely conform to the proposed hypothesis. Using Distributed Alignment Search, we identify causally effective subspaces aligned with the variables in the polarity-composition hypothesis. These representations transfer across negation forms and partially to a subset of MNLI. We release the dataset and code.
| Comments: | Accepted to EACL 2027 (main conference) |
| Subjects: | Computation and Language (cs.CL) |
| Cite as: | arXiv:2611.23185 [cs.CL] |
| (or arXiv:2611.23185v1 [cs.CL] for this version) | |
| https://doi.org/10.48550/arXiv.2611.23185 arXiv-issued DOI via DataCite (pending registration) |
Submission history
From: Nhi Hoai Doan [view email]
[v1]
Tue, 17 Nov 2026 18:30:35 UTC (380 KB)
来源:arXiv cs.CL · arxiv.org