- Rating
- 3(+1)
- Confidence
- 4
- Informativeness
- 2
- Presentation
- 2
- Soundness
- 2
Read review
Paper summary
This paper studies open-set graph anomaly detection, where labeled anomalies cover only seen anomaly classes while the test graph also contains unseen anomaly classes. The main idea is that the unlabeled set already contains unseen anomalies, so it can provide useful supervision without explicitly synthesizing unseen anomaly examples. NoRD combines supervised detection on labeled normal/seen-anomaly nodes with a normal-referenced discrimination objective. Normal-reference normalization and a mixture link are introduced to avoid incorrectly pushing unlabeled normal nodes toward high anomaly scores. Experiments on seven public datasets and an additional large internal payment graph show consistent improvements over a broad set of baselines, including recent open-set GAD methods such as NSReg and DEMO.
Critical strengths
- The motivation is clear. NoRD makes effective use of unlabeled nodes, which may already contain unseen anomalies, rather than relying on synthetic unseen anomalies.
- The normal-reference normalization and mixture link are well aligned with the motivation, helping exploit the unlabeled population while reducing the influence of unlabeled normal nodes.
Critical weaknesses
- The method relies on a representative normal reference.
- The open-set evaluation on Yelp and T-Finance relies on constructed anomaly subclasses. Yelp and T-Finance provide only binary anomaly labels.
- Limited comparisons with recent generalist GAD methods.
- Although the paper includes several open-set GAD baselines, it lacks comparisons or sufficient discussion of recent generalist and zero-shot GAD methods, including ARC [1], UNPrompt [2], IA-GGAD [3], DR-GGAD [4], AnomalyGFM [5], and REFI-GAD [6].
[1] Liu, Y., Li, S., Zheng, Y., Chen, Q., Zhang, C., and Pan, S. ARC: A Generalist Graph Anomaly Detector with In-Context Learning. NeurIPS, 2024. [2] Niu, C., Qiao, H., Chen, C., Chen, L., and Pang, G. Zero-shot Generalist Graph Anomaly Detection with Unified Neighborhood Prompts. IJCAI, 2025. [3] Zhang, X., He, Z., Fu, C., and Xie, C. IA-GGAD: Zero-shot Generalist Graph Anomaly Detection via Invariant and Affinity Learning. NeurIPS, 2025. [4] Fu, C., He, Z., Zhang, X., Xie, C., Jin, X., and Yang, Y. DR-GGAD: Dual Residual Centering for Mitigating Anomaly Non-Discriminativity in Generalist Graph Anomaly Detection. ICLR, 2026. [5] Qiao, H., Niu, C., Chen, L., and Pang, G. AnomalyGFM: Graph Foundation Model for Zero/Few-shot Anomaly Detection. KDD, 2025. [6] Liu, Y., Liu, Y., Zheng, Y., Liew, A. W.-C., Cao, X., and Pan, S. Rethinking Feature Alignment in Generalist Graph Anomaly Detection: A Relational Fingerprint-based Approach. ICML, 2026.
Questions for authors
see weaknesses
4 comments
Response to Reviewer 4TNy (2/2)
W3: Limited comparisons with recent generalist GAD methods.
W4: Although the paper includes several open-set GAD baselines, it lacks comparisons or sufficient discussion of recent generalist and zero-shot GAD methods, including ARC [1], UniPrompt [2], IA-GGAD [3], DR-GGAD [4], AnomalyGFM [5], and REFI-GAD [6].
A3–A4: Thank you for pointing us to this line of work; the comparison turned out to be informative beyond the ranking itself. Following your suggestion, we ran all six methods on our seven public graphs, under the same partitions and metrics as the other baselines. The results are given below and have been added to Tables 1–2 and Appendix C.1.
Table R2. AUROC / AUPRC (%) on all anomalies.
| Method | Photo | Computers | CS | Yelp | T-Finance | ogbn-arxiv | ogbn-proteins | Avg. |
|---|---|---|---|---|---|---|---|---|
| ARC | 57.38 / 9.82 | 53.28 / 15.90 | 50.73 / 22.27 | 64.84 / 24.73 | 76.80 / 25.72 | 48.43 / 16.52 | 53.11 / 9.21 | 57.80 / 17.74 |
| UNPrompt | 49.04 / 6.74 | 54.03 / 16.82 | 68.86 / 41.19 | 45.27 / 13.22 | 59.76 / 23.68 | 56.80 / 20.23 | 73.33 / 21.09 | 58.16 / 20.42 |
| AnomalyGFM | 55.32 / 9.74 | 46.35 / 14.38 | 77.18 / 57.95 | 57.84 / 16.13 | 66.81 / 9.88 | 56.07 / 19.99 | 49.41 / 8.56 | 58.43 / 19.52 |
| IA-GGAD | 46.96 / 8.68 | 63.20 / 35.98 | 70.53 / 38.73 | 58.61 / 20.07 | 69.83 / 19.04 | 57.71 / 20.72 | 68.98 / 54.74 | 62.26 / 28.28 |
| DR-GGAD | 61.27 / 11.25 | 44.94 / 13.66 | 57.21 / 25.97 | 30.22 / 10.26 | 60.98 / 26.54 | 51.88 / 17.14 | 36.48 / 6.11 | 49.00 / 15.85 |
| ReFi-GAD | 74.82 / 45.90 | 54.82 / 17.61 | 68.80 / 36.86 | 72.19 / 36.21 | 88.25 / 54.01 | 52.92 / 17.46 | 79.52 / 60.85 | 70.19 / 38.41 |
| NoRD | 88.05 / 58.41 | 82.29 / 58.19 | 97.23 / 92.73 | 76.77 / 37.89 | 93.26 / 78.05 | 63.25 / 29.57 | 83.59 / 67.96 | 83.49 / 60.40 |
Table R3. AUROC / AUPRC (%) on unseen anomalies.
| Method | Photo | Computers | CS | Yelp | T-Finance | ogbn-arxiv | ogbn-proteins | Avg. |
|---|---|---|---|---|---|---|---|---|
| ARC | 58.10 / 9.41 | 53.38 / 13.60 | 41.59 / 14.78 | 55.62 / 12.98 | 76.39 / 16.26 | 48.41 / 12.98 | 51.38 / 5.00 | 54.98 / 12.14 |
| UNPrompt | 50.10 / 6.23 | 53.99 / 14.47 | 68.67 / 24.56 | 48.60 / 6.95 | 57.25 / 12.46 | 52.87 / 13.76 | 65.55 / 23.33 | 56.72 / 14.54 |
| AnomalyGFM | 55.85 / 9.24 | 46.28 / 12.28 | 80.17 / 35.83 | 46.56 / 8.64 | 67.67 / 5.75 | 56.19 / 13.95 | 53.25 / 5.01 | 58.00 / 12.96 |
| IA-GGAD | 50.70 / 5.33 | 63.21 / 23.68 | 70.38 / 16.74 | 53.20 / 10.48 | 85.60 / 59.24 | 50.68 / 12.82 | 66.97 / 4.39 | 62.96 / 18.95 |
| DR-GGAD | 62.44 / 7.43 | 44.93 / 11.68 | 50.09 / 17.05 | 41.13 / 5.63 | 59.24 / 18.11 | 51.82 / 13.49 | 33.90 / 3.28 | 49.08 / 10.95 |
| ReFi-GAD | 54.10 / 6.29 | 51.65 / 13.48 | 67.39 / 32.57 | 67.89 / 18.41 | 84.62 / 68.32 | 51.89 / 13.34 | 71.47 / 8.89 | 64.14 / 23.04 |
| NoRD | 82.69 / 27.09 | 81.63 / 44.98 | 96.97 / 90.81 | 71.90 / 22.07 | 94.77 / 81.84 | 56.21 / 14.97 | 74.91 / 43.94 | 79.87 / 46.53 |
NoRD achieves the highest mean on every graph and metric. Beyond the ranking, we would like to discuss what we think the comparison shows.
We believe the gap mainly reflects what each family of methods learns from. Generalist detectors are trained on other graphs and are designed to work on a new graph with few or no labels, so their notion of an anomaly comes mostly from elsewhere. The per-graph results suggest that this notion transfers unevenly: the best generalist method changes from graph to graph, and a method that does well on one graph can fall to or below chance level on another. Open-set GAD instead trains on the target graph itself, and the premise of NoRD is that the unlabeled nodes of this graph already contain the unseen anomalies, so the detector can learn what is anomalous here rather than in general. Adapting with a few target labels narrows the gap only in part: ReFi-GAD, which does so, is the strongest of the six on all anomalies, but its advantage over the zero-shot methods largely disappears on unseen ones.We therefore see the two lines of work as addressing different problems, transferring a detector across graphs versus going beyond the labeled anomaly classes within one graph, rather than as direct competitors, and we have added a discussion along these lines to Related Work.
Thank you again for helping us broaden the comparison. We sincerely hope that our responses have addressed your concerns.
score
The authors have addressed my concerns. I will raise my score.
Thank you very much for your positive reassessment! We are pleased that our responses have addressed your concerns. We sincerely appreciate your careful review and constructive suggestions.
Response to Reviewer 4TNy (1/2)
Thank you for your valuable feedback and for your positive assessment of our motivation and design. We have revised the paper based on your comments, with the changes marked in blue. We detail our responses below, and please kindly let us know if they address your concerns.
A1: We agree with this point. NoRD uses the labeled normal nodes as its reference for what "normal" looks like: the discrimination objective asks which unlabeled nodes differ from them. If the labeled nodes cover only part of the normal population, the normal nodes outside that part also differ from the reference, and the objective can no longer separate them from anomalies as cleanly. The dependence you point out is therefore inherent to the design, and the relevant question is how quickly performance degrades as the reference becomes less representative. Our original experiments sampled the labeled normal nodes uniformly and could not answer it. We have now added an experiment (Section 4.2 and Appendix C.4) in which, on Photo, Computers, and CS, 50%, 80%, or 100% of the labeled normal nodes are drawn from a single normal class, with NSReg and DEMO run on the same partitions.
Table R1. Unseen AUROC / AUPRC (%) when the labeled normal nodes are biased toward one normal class, averaged over Photo, Computers, and CS.
What Table R1 shows is that this reliance does not turn into fragility. NoRD stays ahead of NSReg and DEMO at every level of bias, so even a poorly chosen reference never leaves it behind methods that use the labeled normal nodes differently. A reference that is skewed but still contains the other normal classes remains quite useful: at 50%, NoRD's lead in unseen AUROC over the stronger baseline widens from 1.7 to 4.3 points, and its accuracy (83.36) is close to what the two baselines reach with an unbiased reference (84.79 and 85.41). The hard case is 100%, where the other normal classes are missing from the reference altogether. Accuracy drops considerably there, but it does so for all three methods, and NoRD keeps a narrow lead. We therefore see the remaining loss as a difficulty that unrepresentative normal labels pose for open-set GAD in general, rather than one specific to normal-referenced discrimination.
A2: Open-set evaluation needs anomaly subclasses, so that some of them can be held out as unseen, but Yelp and T-Finance come with binary labels only. For these two graphs we follow the protocol of NSReg (ICLR 2025, Appendix B.2), which clusters the anomalous nodes with k-means on Deep Graph Infomax representations, so that open-set methods can be compared under a common protocol; Section 4.1 now states this construction and its source explicitly.