LLM品牌推荐中的系统归因：单条回复可识别系统，聚合品牌画像无法迁移
原标题：System Attribution in LLM Brand Recommendations: Single Responses Identify the System, Aggregated Brand Profiles Do Not Transfer
AI 摘要
该论文研究LLM品牌推荐中的系统归因问题，使用包含6,475条存储回复的语料库（来自五个部署端点，覆盖礼物推荐、企业声誉和品类归属查询）。研究发现，单条回复可通过字符n-gram分类器以97.84%准确率识别其来源系统，主要依赖答案的表面格式而非品牌内容；但将回复聚合为模型-领域-条件单元后，行为特征跨领域无法迁移，在一个领域训练的模型会将另一领域的全部单元错误归类。结论是答案的表面形式携带系统信号，而聚合品牌行为不能跨查询领域泛化。
正文节选
System Attribution in LLM Brand Recommendations: Single Responses Identify the System, Aggregated Brand Profiles Do Not Transfer Abstract Audits of AI visibility summarise the brand recommendations of deployed language models into per-system profiles. This paper asks whether such a profile describes the system, and puts the question to one corpus at two units of observation. The corpus holds 6,475 stored responses, 6,324 of them analysable, collected between December 2025 and February 2026 from