研究：监督式成对图表示的局部证据、混淆因素与事后校正
Interpreting Supervised Pairwise Graph Representations: Local Evidence, Confounding, and Provable Post-hoc Correction
一篇冷静的方法论论文：用 ROC-AUC 0.839 对 0.559 这些实验数据说明图表示的优势可能只是混淆变量，做图模型解释的同行可以看看。
arXiv 论文在受控 Transformer 错误检测任务中测得监督式成对图表示 ROC-AUC 为 0.839，超过 attention-as-adjacency 的 0.559。但在控制序列长度后（该变量可从池化隐藏状态以 R²=0.9995 恢复），其对模型置信度的表观优势消失。在 HCP Emotion 集成图上，拓扑感知增强对 ROC-AUC 的改动不超过 0.004，三层 GCN 相对带符号节点汇总基线无可测提升。论文提出的 Fisher 事后校正器在各特征表示下满足分布无关保证，并在 fMRI 校准场景中正确给出统计无信息的界。
Interpreting Supervised Pairwise Graph Representations: Local Evidence, Confounding, and Provable Post-hoc Correction
Supervised pairwise graph representations convert pairwise predictive relations into sample-specific graphs, but graph-level accuracy does not by itself show that higher-order topology carries the signal. In a controlled Transformer error-detection task, supervised pairwise graphs outperform attention-as-adjacency (ROC-AUC $0.839$ versus $0.559$; the comparable no-confidence pair is $0.687$ versus $0.559$), but their apparent advantage over model confidence disappears after controlling for sequence length, which is almost perfectly recoverable from the pooled hidden state ($R^2=0.9995$). On published HCP Emotion ensemble graphs, signed node aggregation already acts as an ensemble of supervised pairwise predictions. Edge magnitude alone is near chance; topology-aware augmentations change ROC-AUC by at most $0.004$, with only sign-separated degree showing small positive changes across all tested graph sizes, while a validation-tuned three-layer GCN provides no measurable improvement over the signed node-summary baseline. A Fisher post-hoc corrector satisfies its distribution-agnostic guarantees across the evaluated feature representations and remains stable under reduced labelled budgets; in a near-error-free fMRI calibration regime the bound correctly becomes statistically uninformative. A separate cross-attention analysis identifies a redundant reversal circuit and downstream EOS-timing and copy failures. Higher-order topology should therefore be credited only after local supervised evidence and recoverable nuisance variables have been excluded.