Research map / Graph learning / Graph self-supervised learning
Rethinking Feature Alignment in Generalist Graph Anomaly Detection: A Relational Fingerprint-based Approach
Yujing Liu, Yixin Liu, Yu Zheng, Alan Wee-Chung Liew, Xiaofeng Cao, Shirui Pan
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| replace PCA feature alignment with relational fingerprint (REFI) (ablation) | REFI-GAD · Group 2 datasets (Cora, Flickr, Facebook, etc.) | AUROC (%): Flickr 71.71 -> 89.61; Cora 53.28 -> 96.14 | Improved |
| add SNR-guided domain-adaptive refinement (ablation) | REFI-GAD · Group 2 datasets | AUROC (%): Flickr 88.08 -> 89.61; Facebook 93.30 -> 94.67; average improvement of 0.70% | Improved |
| add degree dimension to REFI (ablation) | REFI-GAD · Flickr | AUROC (%): 79.78 -> 89.61 | Improved |
| add global directional consistency to REFI (ablation) | REFI-GAD · Flickr | AUROC (%): 82.05 -> 89.61 | Improved |
| add global directional consistency to REFI (ablation) | REFI-GAD · Cora | AUROC (%): 96.62 -> 96.14 | No clear effect |
| add all five REFI attributes (any single one removed lowers average) (ablation) | REFI-GAD · 7 Group 2 datasets average | AUROC (%): 80.52-82.54 (single removed) -> 83.41 | Improved |
| increase context size k up to 50 (ablation) | REFI-GAD · Citeseer, Weibo, Blog, CS, Photo | AUROC (%) | Improved |
| pre-train on source graphs (vs training-free) | REFI-GAD · 14 datasets | AUROC gain: avg +4.67%, 12 of 14 positive | Improved |
| pre-train on source graphs (vs training-free) | ARC, AnomalyGFM, UNPrompt, IA-GGAD · 14 datasets | AUROC gain: ARC +1.20%, IA-GGAD +1.13%, UNPrompt -0.92%, AnomalyGFM -0.12% | Mixed |
| fully cross-domain source training (citation as target) | existing generalist GAD baselines · Cora, Citeseer, ACM, CS, Pubmed | AUROC gain: AnomalyGFM -10.79%, UNPrompt -6.33%, ARC -0.11%, IA-GGAD -0.21% | Hurt |
| fully cross-domain source training with REFI | REFI-GAD · Cora, Citeseer, ACM, CS, Pubmed | AUROC gain: +4.15% average | Improved |
| remove neighborhood consistency modules from training-free ARC/UNPrompt (ablation) | ARC, UNPrompt (training-free) · 14 datasets | average AUROC | Improved |
Closest papers by meaning
- ARC: A Generalist Graph Anomaly Detector with In-Context Learning (NeurIPS 2024 · similarity 0.90)
- Open-Set Graph Anomaly Detection via Normal Structure Regularisation (ICLR 2025 · similarity 0.87)
- Graph Anomaly Detection as Dynamical Transport: Training-Free Scoring via Empirical Bayes (NeurIPS 2026 · similarity 0.87)
- UniGAD: Unifying Multi-level Graph Anomaly Detection (NeurIPS 2024 · similarity 0.86)
- UniFORM: Towards Unified Framework for Anomaly Detection on Graphs (AAAI 2025 · similarity 0.86)
- Generative Semi-supervised Graph Anomaly Detection (NeurIPS 2024 · similarity 0.86)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.
