发表机构
Posts and Telecommunications Institute of Technology; The University of Sydney; Van Lang University（邮政与电信技术学院; 悉尼大学; 万朗大学）
机构由 AI 辅助整理，请以论文原文为准。AI 中文总结
本文首次对表格基础模型的无训练节点分类进行可靠性研究，提出同质性门控扩散分数HG-DAPS，在保持共形覆盖有效的同时减小预测集大小，并揭示原始同质性门控在类别不平衡图中的覆盖率陷阱。
AI 中文摘要
表格基础模型（TFMs）可以通过将节点和邻域特征作为表格行与标记的上下文行一起读取，从而在不进行训练的情况下对图节点进行分类。该方向的工作报告了预测性能，而未涉及共形覆盖率或预测集大小。据我们所知，我们首次对该设置进行了可靠性研究，使用TabICL作为TFM，并将每个图的一半作为标记上下文。对于任何在校准前固定的预测器，冻结的上下文内预测器使得分裂共形预测在有限样本中精确有效，无需训练、验证折或在目标图上的调参。对十个图的审计显示，无训练的TabICL后验在九个图上的期望校准误差（ECE）低于带温度缩放的GCN（GCN+TS）。其在十个图上的平均ECE为0.019，比GCN+TS的0.029低约35%。我们还引入了HG-DAPS，一种无训练的扩散分数，其同质性门仅读取上下文标签，因此保证仍然成立。相对于自适应预测集（APS），它在六个同质图上将平均集大小减少了5.8%至17.1%，在四个异质图上的变化小于1%。在两个二分类、类别不平衡的图上，一个预先注册的陷阱案例表明，对原始同质性而非调整后的同质性进行门控，使得低同质性节点的覆盖率降低了0.27和0.12。边际覆盖率保持在名义上的0.90，掩盖了这一下降。
英文摘要
Tabular foundation models (TFMs) can classify the nodes of a graph without training on it, by reading node and neighborhood features as table rows next to labeled context rows. Work in this line reports predictive performance, not conformal coverage or prediction-set size. To our knowledge, we give the first reliability study of the setting, with TabICL as the TFM and half of each graph as labeled context. As for any predictor fixed before calibration, a frozen in-context predictor makes split conformal prediction exactly valid in finite samples, with no training, validation fold, or tuning on the target graph. An audit across ten graphs then shows that the training-free TabICL posterior has lower expected calibration error (ECE) than GCN with temperature scaling (GCN+TS) on nine of them. Its mean ECE over the ten graphs is 0.019, about 35 percent below the 0.029 of GCN+TS. We also introduce HG-DAPS, a training-free diffusion score whose homophily gate reads only the in-context labels, so the guarantee still holds. Relative to adaptive prediction sets (APS), it reduces mean set size by 5.8 to 17.1 percent on six homophilous graphs and changes it by under 1 percent on four heterophilous ones. On two binary, class-imbalanced graphs, a pre-registered trap case shows that gating on raw rather than adjusted homophily lowers coverage among low-homophily nodes by 0.27 and 0.12. Marginal coverage stays at the nominal 0.90 and masks this drop.
Comments6 pages, 4 figures, 1 table
