发表机构
National Taiwan University（国立台湾大学）
机构由 AI 辅助整理，请以论文原文为准。AI 中文总结
本研究揭示多语言大模型评估中翻译同构假设的缺陷，提出α-不可对齐性概念，并通过行为、机制和诊断证据表明某些文化概念无法跨语言对齐，强调尊重文化差异是多语言AI的前提。
AI 中文摘要
当前对多语言大语言模型（LLMs）的评估依赖于一个隐含的翻译同构假设（TIA）：即不同语言间的语义结构是一致的，且可以在不损失信息的情况下相互映射。我们认为，这一假设不仅在实践中被违反，而且在原则上对于一类类型学上可识别的概念（包括语用标记、敬语和历时分层术语）是不适定的。我们使用用法云框架形式化了这一失败，将概念表示为上下文嵌入的点集。我们将α-不可对齐性定义为任何同时保持词汇保真度（质心对应）和结构保真度（局部邻域拓扑）的映射的不可能性。我们提供了三层证据。在行为层面，我们展示了FLORES-200翻译失败可由语言家族和资源类别预测，但不由文字系统预测，且LOBSTER推理得分因家族而异。在机制层面，我们报告了关于雅美语（Yami）的九模型案例研究中的表征-干预差距（RIG）：模型的激活编码了一种规律性，沿此规律雅美语与其他低资源和南岛语系语言分组，然而对语言特定神经元的干预并未显示出优于随机掩码的明显优势：该规律性可见但无法被此干预利用。最后，我们将这些发现操作化为一个多维诊断画像：循环一致性、语用负荷分歧、流形曲率不匹配和RIG。我们认为，将文化能力压缩为单一标量会激励“概率扁平化”，而识别不可对齐类别是AI尊重而非抹除文化差异的前提。这表明多语言对齐不是一个单一明确定义的目标，而是一组相互不兼容的投影。
英文摘要
Current evaluation of multilingual Large Language Models (LLMs) rests on an implicit Translation-Isomorphism Assumption (TIA): that semantic structures across languages are congruent and mutually mappable without loss of information. We argue that this assumption is not merely violated in practice, but ill-posed in principle for a typologically identifiable class of concepts, including pragmatic markers, honorifics, and diachronically stratified terms. We formalize this failure using a usage-cloud framework, representing concepts as point sets of contextualized embeddings. We define $α$-unalignability as the impossibility of any mapping that simultaneously preserves lexical faithfulness (centroid correspondence) and structural faithfulness (local neighborhood topology). We provide three layers of evidence. Behaviorally, we show that FLORES-200 translation failures are predicted by language family and resource class but not by script, and that LOBSTER reasoning scores vary by family. Mechanistically, we report a Representation-Intervention Gap (RIG) in a nine-model case study on Yami: the models' activations encode a regularity along which Yami groups with other low-resource and Austronesian languages, yet interventions on language-specific neurons show no demonstrated advantage over random masks: the regularity is visible but not usable by this intervention. Finally, we operationalize these findings into a multidimensional diagnostic profile: Cycle-Consistency, Pragmatic-Load Disagreement, Manifold-Curvature Mismatch, and RIG. We argue that collapsing cultural competence into a single scalar incentivizes "probabilistic flattening," and that recognizing the unalignable class is a precondition for AI that respects, rather than erases, cultural divergence. This suggests that multilingual alignment is not a single well-defined objective, but a set of mutually incompatible projections.
CommentsPosition paper. 32 pages (10 pages main text), 6 figures, 12 tables
