知晓并非陈述:跨边共同体在 LLM 表征中的身份固化(Identity Ossification)
Knowing Is Not Telling: Identity Ossification in LLM Representations of Trans-border Communities
Minglu Sun · Lan Chen · Hang Jiang
中文摘要
民族与文化身份很少与政治边界吻合,移民与长期存在的跨境联系使这一张力更为突出。Sami、Rohingya、Kurdish 等共同体长期分居于多个国家之上。这一张力在中国西南与中南半岛(Mainland Southeast Asia)的交界处尤为明显,大量共同体跨越国界共享历史、语言与文化。研究考察中国西南边境沿线的九个跨境民族共同体,分析大语言模型(LLMs)如何表征这些身份。论文定义并研究"身份固化"(identity ossification):即 LLM 将流动的、跨国的民族身份坍缩为单一的、固定的民族国家身份的模式。提出跨境身份偏差诊断框架(Transborder Identity Ossification Diagnostic Framework, TIODF),通过对比直接的事实性问题与开放式叙事性问题,检验固化现象究竟源于知识缺失还是模型答案的叙事框架。在英文与中文下测试 GPT-5.1 与 DeepSeek-V3.2。固化程度因模型与查询语言不同而明显不同——在 GPT-5.1 的英文回答中仅出现 10.1%,而在 DeepSeek-V3.2 的中文回答中高达 62.6%。模型即便在事实问题上回答正确,仍常常以单一民族国家视角讲述故事。这种解耦表明固化反映的是叙事框架而非简单的知识缺失,对多语言 AI 系统中文化公平性的评估具有重要意义。
关键要点
- 01研究关注中国西南与中南半岛交界的九个跨境民族共同体,检验 LLM 对其身份的表征是否存在系统性偏差。
- 02提出"身份固化"(identity ossification)概念,即模型将跨境流动的民族身份坍缩为单一固定的国家身份。
- 03设计跨境身份偏差诊断框架 TIODF,通过事实性问题与叙事性问题的对比,区分知识缺失与叙事框架两类成因。
- 04在 GPT-5.1 与 DeepSeek-V3.2 的中英文测试中,固化比例差异显著,DeepSeek-V3.2 中文回答中达 62.6%,GPT-5.1 英文回答仅 10.1%。
- 05即使事实问答正确,模型仍倾向于单一国家叙事,说明固化更多源自叙事框架而非知识缺失。
解读
尚无解读。
原始英文摘要
arXiv:2610.12610v1 Announce Type: cross Abstract: Ethnic and cultural identities rarely align with political borders, a tension made sharper by migration and long-standing cross-border ties. Communities such as the Sami, the Rohingya, and the Kurds have long lived across more than one state. This tension is especially clear in mainland Southeast Asia, where many communities share history, language, and culture across state lines. We study nine such transborder ethnic communities along China's southwestern border with mainland Southeast Asia, looking at how large language models (LLMs) represent them. We define and study identity ossification: a pattern where LLMs turn a fluid, cross-border ethnic identity into a single, fixed nation-state identity. We introduce the Transborder Identity Ossification Diagnostic Framework (TIODF) that compares direct factual questions with open-ended narrative questions to test whether ossification comes from missing knowledge or from how models frame their answers. We test GPT-5.1 and DeepSeek-V3.2 in English and Chinese. Ossification varies substantially across models and query languages. It appears in only 10.1% of GPT-5.1's English answers, but in 62.6% of DeepSeek-V3.2's Chinese answers. Models often still tell a single-nation story even when they answer the factual question correctly. This decoupling suggests that ossification reflects narrative framing rather than simple knowledge absence, which matters for how we test cultural fairness in multilingual AI systems.