From Subtle to Significant: Prompt-Driven Self-Improving Optimization in Test-Time Graph OOD Detection
Luzhi Wang, Xuanshuo Fu, He Zhang, Chuang Liu, Xiaobao Wang, Hongbo Liu · 2026
Read From Subtle to Significant: Prompt-Driven Self-Improving Optimization in Test-Time Graph OOD Detection by Luzhi Wang, Xuanshuo Fu, He Zhang, Chuang Liu, Xiaobao Wang, Hongbo Liu (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-2602-17342.txt.
Description
Luzhi Wang led a six-author team lifting graph out-of-distribution detection at test time. Such checks flag test graphs that deviate from the training mix, a safeguard for graph neural networks deployed in open-world settings. Recent test-time training methods avoid supervisory data but run one-pass inference, so they cannot progressively correct errors to amplify out-of-distribution signals. The paper proposes SIGOOD, an unsupervised framework joining continuous self-learning with test-time training.
SIGOOD generates a prompt to build a prompt-enhanced graph that amplifies potential OOD cues. An Energy Preference Optimization loss tunes the prompt using energy variations between the original and boosted graphs, and iterative tuning folds the prompt into the detection model in a self-lifting loop. The final boosted graph drives the verdict. Evaluations on twenty-one real-world datasets confirm the method's effectiveness and outperformance.
Catalog record
| CREATOR | Luzhi Wang, Xuanshuo Fu, He Zhang, Chuang Liu, Xiaobao Wang, Hongbo Liu |
|---|---|
| YEAR | 2026 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 39 KB |
| VERSION | 1.0 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2602-17342