用LLM特征拼接反而降低图神经网络性能，尤其在同质图上。
LLM Features Can Hurt GNNs: Concatenation Interference on Homophilous Graph Benchmarks
- 将LLM特征直接拼接输入会干扰图神经网络训练
- 在PubMed上准确率下降17.0个百分点，其他数据集也普遍下降
- 特征可区分性低时拼接有害，适合研究特征融合的边界条件
将LLM生成的节点特征通过纯拼接方式加入图神经网络（GNNs），常被报道能提升标准基准上的精度。我们发现相反现象：在同质图基准上，这种拼接方式系统性地降低精度，而端到端的LLM方法却有效。以MLP为骨干，在Planetoid公开划分和词袋原始特征下，拼接SBERT编码的GPT-4o-mini TAPE特征导致PubMed测试准确率下降-17.0±0.3个百分点，Cora下降-4.3±0.6个百分点（CiteSeer下降-0.6±0.8个百分点，处于种子噪声范围内）。当放宽条件（如使用GCN/GCNII/GAT、随机划分、更小编码器）后，下降幅度减弱，并在中等同质性数据集WikiCS（+4.4 pp）和ogbn-arxiv（+11.7 pp）逆转。我们提出一个简单度量Δ_sig（LLM单独判别能力），其与拼接代价相关性（r²=0.38）高于同质性（r²=0.06，N=9，置信区间重叠）。最佳变化点τ=13.8个百分点，“Δ_sig ≤ τ”预测非正拼接成本的规则在9个数据集中正确分类7个；因60%的样本中τ落在[5,30]之间，故Δ_sig作为解释工具而非精确筛选器。在PubMed的维度控制消融实验中，特征下降介于同源PCA（-2.3 pp）与同维高斯噪声（-37.3 pp）之间，排除了维度和权重衰减的干扰。九种PubMed配置符合幂律关系|Δ_concat| ∝ (sqrt(d_l/n))^1.31，r²=0.97；低Δ_sig、小样本区域正是-17.0百分点显著下降的位置。
原文摘要 · Abstract (English)
Adding LLM-generated node features to graph neural networks (GNNs) is widely reported to improve accuracy on standard benchmarks. We document a contrasting observation: when LLM features are introduced through pure input concatenation (rather than joint training, distillation, or prompt-conditioning), they can systematically degrade accuracy on the same homophilous benchmarks where end-to-end LLM pipelines succeed. With an MLP backbone on the Planetoid public split and bag-of-words original features, concatenating SBERT-encoded GPT-4o-mini TAPE features reduces PubMed test accuracy by -17.0 +/- 0.3 pp and Cora by -4.3 +/- 0.6 pp (CiteSeer -0.6 +/- 0.8 pp, within seed noise). The drop attenuates as we relax each condition (GCN / GCNII / GAT backbones, random splits, smaller encoders) and reverses on medium-homophily WikiCS (+4.4 pp) and ogbn-arxiv (+11.7 pp). To predict when concatenation helps versus hurts, we report a simple measure of LLM-alone discriminability, Delta_sig. Across 9 datasets Delta_sig correlates with the concatenation cost more strongly than homophily at point estimate (r^2 = 0.38 vs. 0.06; N=9, bootstrap CIs overlap). The bootstrap-best change-point is tau = 13.8 pp, and the rule "Delta_sig <= tau predicts non-positive concat cost" classifies 7/9 datasets correctly; since 60% of bootstrap samples place tau in [5, 30] pp, we treat Delta_sig as an interpretive lens rather than a precision filter. A dimension-controlled ablation on PubMed places the LLM-feature drop between same-source PCA (-2.3 pp) and same-dim Gaussian noise (-37.3 pp), ruling out dimensionality and weight-decay artifacts. Nine PubMed configurations fit a power law |Delta_concat| proportional to (sqrt(d_l/n))^1.31 with r^2 = 0.97; the low-Delta_sig, small-n corner is exactly where the headline -17 pp PubMed deficit appears.
Thank you to arXiv for use of its open access interoperability. PaperDance 不是 arXiv 官方产品；中文卡片由大模型生成，请以原文为准。