arXiv cs.AI· Tony Mason·· 9 天前AI 评分24
社会语域如何塑造大语言模型的指令拓扑：祈使语气导致跨语言行为反转
Imperative Interference: Social Register Shapes Instruction Topology in Large Language Models
AI 导读
研究发现，同一语义的系统提示词指令在英语中协同、在西班牙语中却相互竞争，指令拓扑出现反转。对四种语言、四个模型的指令级消融实验显示，这一现象由社会语域介导：将单个指令块改写为陈述句可使跨语言方差降低 81%（p = 0.029，置换检验）；改写 11 个祈使块中的 3 个，即可将西班牙语指令拓扑从竞争转为协同，并外溢影响未改写块。
正文 · 原文
Abstract:System prompt instructions that cooperate in English compete in Spanish, with the same semantic content, but opposite interaction topology. We present instruction-level ablation experiments across four languages and four models showing that this topology inversion is mediated by social register: the imperative mood carries different obligatory force across speech communities, and models trained on multilingual data have learned these conventions. Declarative rewriting of a single instruction block reduces cross-linguistic variance by 81% (p = 0.029, permutation test). Rewriting three of eleven imperative blocks shifts Spanish instruction topology from competitive to cooperative, with spillover effects on unrewritten blocks. These findings suggest that models process instructions as social acts, not technical specifications: "NEVER do X" is an exercise of authority whose force is language-dependent, while "X: disabled" is a factual description that transfers across languages. If register mediates instruction-following at inference time, it plausibly does so during training. We state this as a testable prediction: constitutional AI principles authored in imperative mood may create language-dependent alignment. Corpus: 22 hand-authored probes against a production system prompt decomposed into 56 blocks.
| Subjects: | Computation and Language (cs.CL); Artificial Intelligence (cs.AI); Software Engineering (cs.SE) |
| Cite as: | arXiv:2603.25015 [cs.CL] |
| (or arXiv:2603.25015v2 [cs.CL] for this version) | |
| https://doi.org/10.48550/arXiv.2603.25015 arXiv-issued DOI via DataCite |
Submission history
From: Tony Mason [view email]
[v1]
Thu, 26 Mar 2026 04:25:14 UTC (56 KB)
[v2]
Wed, 7 Oct 2026 18:16:09 UTC (53 KB)
来源：arXiv cs.AI · arxiv.org