AI Paper Atlas

Research map / Graph learning / Graph self-supervised learning

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

AAAI 2026 · Graph learning › Graph self-supervised learning · approach: Graph out-of-distribution generalization · 0 citations (OpenAlex)

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 changedWhereMeasuredEffect
add the energy preference optimization (EPO) loss (ablation)SIGOOD · BZR/COX2, PTC-MR/MUTAG, Esol/MUVAUCImproved
use an MLP prompt generator instead of a free parameter matrix (ablation)SIGOOD · BZR/COX2, PTC-MR/MUTAG, Esol/MUVAUCImproved
increase prompt generator depth to 3 layers from 1-2 (ablation)SIGOOD · BZR/COX2, PTC-MR/MUTAG, Esol/MUVAUCImproved
increase self-improving iterations from 1000 to 10000 (ablation)SIGOOD · Esol/MUVAUC: peak 87.72%Improved
set beta to a moderate value (80) (ablation)SIGOOD · BZR/COX2AUC: best 87.36%Improved
SIGOOD vs best test-time baseline GOODATtest-time graph OOD detection · ClinTox/LIPOAUC: 62.46 -> 71.33Improved
SIGOOD vs GOODAT on anomaly detectiongraph anomaly detection · DD and p53AUC: DD 77.62 -> 72.59; p53 63.27 -> 60.51Hurt
SIGOOD vs GOODAT on anomaly detectiongraph anomaly detection · ENZYMESAUC: 52.33 -> 76.80Improved

Earlier methods this paper compares with

Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 1 on most shared results.

Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.

Closest papers by meaning

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.

Open scime draft

Get the next scime Dispatch and a monthly field update

Free. We send a confirmation link first; every email has a one-click unsubscribe. Privacy