Graph Anomaly Detection on Citeseer (F1, Avg. Rank)
87.5Anomaly F1N2NSC
Evaluation results
| Method | Links | ||
|---|---|---|---|
| N2NSCLLM Backbone=Qwen3-8B, PEFT=LoRA (rank 64), GNN Encoder=2-layer GAT, GNN heads=8, GNN output dimension=256, Contrastive selection K=22026.06 | 87.5 | 1 | |
| GraphSAGECategory=GNN-based methods2026.06 | 63.7 | 5.5 | |
| BWGNNCategory=GNN-based methods2026.06 | 61.54 | 8 | |
| GINCategory=GNN-based methods2026.06 | 58.54 | 7 | |
| TAPECategory=Integrating LLMs with graphs2026.06 | 58.54 | 5.25 | |
| GCNCategory=GNN-based methods2026.06 | 55.81 | 7.25 | |
| XGBGraphCategory=GNN-based methods2026.06 | 52.89 | 5 | |
| RFGraphCategory=GNN-based methods2026.06 | 47.06 | 7.62 | |
| PC-GNNCategory=GNN-based methods2026.06 | 42.42 | 7.25 | |
| GHRNCategory=GNN-based methods2026.06 | 38.89 | 9.25 | |
| GATCategory=GNN-based methods2026.06 | 34.48 | 9.62 | |
| CoLLCategory=Integrating LLMs with graphs2026.06 | 25.07 | 14.62 | |
| UNPromptCategory=Integrating LLMs with graphs2026.06 | 23.18 | 15.12 | |
| PMPCategory=GNN-based methods2026.06 | 19.44 | 12.38 | |
| GAAPCategory=GNN-based methods2026.06 | 18.95 | 12.25 | |
| GraphGPTCategory=Integrating LLMs with graphs2026.06 | 14.78 | 17.75 | |
| ConsisGADCategory=Integrating LLMs with graphs2026.06 | 13.48 | 14.12 | |
| InstructGLMCategory=Integrating LLMs with graphs2026.06 | 9.21 | 12 |