Large language models (LLMs) exhibit pronounced social biases. Output-level or data-optimization--based debiasing methods cannot fully resolve these biases, and many prior works have shown that biases are embedded in internal representations. We propose \underline{U}nified \underline{G}raph \underline{I}somorphism for \underline{D}ebiasing large language models (\textit{\textbf{UGID}}), an internal-representation--level debiasing framework for large language models that models the Transformer as a structured computational graph, where attention mechanisms define the routing edges of the graph and hidden states define the graph nodes. Specifically, debiasing is formulated as enforcing invariance of the graph structure across counterfactual inputs, with differences allowed only on sensitive attributes. \textit{\textbf{UGID}} jointly constrains attention routing and hidden representations in bias-sensitive regions, effectively preventing bias migration across architectural components. To achieve effective behavioral alignment without degrading general capabilities, we introduce a log-space constraint on sensitive logits and a selective anchor-based objective to preserve definitional semantics. Extensive experiments on large language models demonstrate that \textit{\textbf{UGID}} effectively reduces bias under both in-distribution and out-of-distribution settings, significantly reduces internal structural discrepancies, and preserves model safety and utility.

翻译：摘要：大语言模型（LLMs）表现出显著的社会偏见。基于输出层或数据优化的去偏方法无法完全消除这些偏见，且诸多先前研究表明偏见已嵌入模型内部表征中。我们提出面向大语言模型去偏的统一图同构方法（\textit{\textbf{UGID}}）——一种内部表征层面的去偏框架，该框架将Transformer建模为结构化计算图，其中注意力机制定义图的路径边，隐藏状态定义图节点。具体而言，去偏被形式化为强制干预输入下的图结构不变性，仅允许敏感属性存在差异。\textit{\textbf{UGID}}在偏见表征区域联合约束注意力路由与隐藏表征，有效阻止偏见跨架构组件迁移。为在不降低通用能力的前提下实现有效的行为对齐，我们引入对敏感logits的对数空间约束以及选择性锚点目标函数以保留定义性语义。在大语言模型上的广泛实验表明，\textit{\textbf{UGID}}在分布内与分布外场景下均能有效降低偏见，显著减少内部结构差异，并保持模型的安全性与实用性。