AI Paper Atlas

Research map / Graph learning / GNN architectures

GraphTextack: A Realistic Black-Box Node Injection Attack on LLM-Enhanced GNNs

Jiaji Ma, Puja Trivedi, Danai Koutra

AAAI 2026 · Graph learning › GNN architectures · approach: Robustness, privacy, and fairness of GNNs · 1 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
inject nodes with jointly optimized structure and text via evolutionary search (GraphTextack) rather than best uni-modal text attackOne-for-all · Cora, r=0.03post-attack accuracy: 74.36 -> 65.75Improved
GraphTextack vs best prior baselines across budgetsOne-for-all · ogbn-products, r=0.01post-attack accuracy: 78.99 vs GANI 78.93No clear effect
GraphTextack vs WTGIA on WikiCS at r=0.05One-for-all · WikiCS, r=0.05post-attack accuracy: 61.35 vs 60.95No clear effect
add crossover to the evolutionary search (ablation)GraphTextack · WikiCS, r=0.05post-attack accuracy: 64.42 -> 61.35Improved
add mutation to the evolutionary search (ablation)GraphTextack · WikiCS, r=0.05post-attack accuracy: 65.19 -> 61.35Improved
add local prediction shift term to fitness (ablation)GraphTextack · WikiCS, r=0.05post-attack accuracy: 66.83 -> 61.35Improved
add PageRank influence term to fitness (ablation)GraphTextack · WikiCS, r=0.05post-attack accuracy: 61.97 -> 61.35No clear effect
add PageRank influence term to fitness (ablation)GraphTextack · WikiCS, r=0.01post-attack accuracy: 71.92 -> 71.69No clear effect
query target model directly instead of surrogate modelsGraphTextack · runtime per injection (ogbn-arxiv, ogbn-products)runtimeImproved
uni-modal attacks (text-only) on LLM-enhanced GNNsLLM-enhanced GNNs · node classificationaccuracy drop: less than 5% dropHurt

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