Dimensions overall score 7.0

GitHub Code Pulse

No public code linked for this paper yet.

• A robustness assessment framework is proposed to systematically evaluate LLM-enhanced GNNs under poisoning attacks.method 95%
• The evaluation includes 24 victim models formed by combining eight LLM/LM-based feature enhancers with three GNN backbones.method 95%
• The framework incorporates six structural poisoning attacks and three textual poisoning attacks at character, word, and sentence levels.method 95%
• Four real-world datasets are used, including one released after the emergence of LLMs, to avoid potential ground truth leakage.method 95%
• LLM-enhanced GNNs exhibit significantly higher accuracy and lower Relative Drop in Accuracy (RDA) than a shallow embedding-based baseline across various attack settings.result 90%
• Key factors contributing to the robustness of LLM-enhanced GNNs include the effective encoding of structural and label information in node representations.result 85%
• The proposed framework provides the most comprehensive coverage of victim models and attack types to date, and incorporates post-LLM dataset evaluation.method 90%
• A robustness assessment framework is proposed to systematically evaluate LLM-enhanced GNNs under poisoning attacks.method 95%
• The framework assesses 24 victim models by combining eight LLM/LM-based feature enhancers with three GNN backbones.method 95%
• The evaluation includes six structural poisoning attacks and three textual poisoning attacks at character, word, and sentence levels.method 95%
• Four real-world datasets are used for evaluation, including one released after the emergence of LLMs to avoid ground truth leakage.method 95%
• LLM-enhanced GNNs exhibit significantly higher accuracy and lower Relative Drop in Accuracy (RDA) than a shallow embedding-based baseline across various attack settings.result 90%

Competitive landscape

Competitor map is still being generated for this paper. Enable generation or check back soon.

Startup potential card