Node-to-Neighborhood Semantic Consistency: Text-Topology Alignment for TAGs Anomaly Detection
Bochen Lin, Jianxiang Yu, Jiayi Wu, Lin Qi, Huang Lu, Xiang Li · 2026
Read Node-to-Neighborhood Semantic Consistency: Text-Topology Alignment for TAGs Anomaly Detection by Bochen Lin, Jianxiang Yu, Jiayi Wu, Lin Qi, Huang Lu, Xiang Li (2026) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2606-30009.txt.
Description
Bochen Lin, Jianxiang Yu, Jiayi Wu, Lin Qi, Huang Lu, and Xiang Li hunt anomalies in text-attributed graphs. In such networks, each node carries a piece of text. Paper abstracts and product descriptions are typical examples. Detection supports fraud checks and academic integrity reviews. A node can look normal alone yet clash with its neighborhood. Earlier methods largely missed that relational signal.
The authors recast detection as a semantic consistency problem. A node is checked against the text and structure around it. Trouble can come from a wording mismatch or structural deviation. Their framework, N2NSC, fuses information along two complementary paths. A language model can then use both wording and local structure. Across eight datasets, N2NSC consistently beats the strongest existing methods. The authors also release their code publicly.
Catalog record
| CREATOR | Bochen Lin, Jianxiang Yu, Jiayi Wu, Lin Qi, Huang Lu, Xiang Li |
|---|---|
| YEAR | 2026 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 47 KB |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2606.30009