Research map / Graph learning / GNN architectures
DuoKD: Dual Knowledge Distillation from Large Language Models for Robust Graph Neural Networks
Cuiying Huo, Xiaotong Huang, Dongxiao He, Yixuan Du, Wenhuan Lu, Di Jin
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 changed | Where | Measured | Effect |
|---|---|---|---|
| add positive knowledge supervision (ablation) | DuoKD · Cora | accuracy: 5.8% drop when removed | Improved |
| add negative knowledge supervision (ablation) | DuoKD · ogbn-arxiv | accuracy | Improved |
| add rationale (semantic and structural) supervision (ablation) | DuoKD · Cora/PubMed/ogbn-arxiv | accuracy | Improved |
| use DuoKD with GraphSAGE student instead of GLEM (SAGE) | DuoKD · PubMed | accuracy: 89.30 -> 88.05 | Hurt |
| use DuoKD instead of LinguGKD (SAGE) | DuoKD · Cora | accuracy: 90.91 -> 91.85 | Improved |
| use DuoKD instead of LinguGKD (GAT) | DuoKD · ogbn-arxiv | accuracy: 77.62 -> 80.13 | Improved |
| use DuoKD instead of GLEM under 40% structural attack | DuoKD · Cora | accuracy: 80.78 -> 84.21 | Improved |
| use DuoKD instead of GLEM under 40% attribute corruption | DuoKD · PubMed | accuracy: 83.30 -> 84.90 | Improved |
| set rationale filter alpha 0.2-0.4 and threshold tau 0.7-0.9 (ablation) | DuoKD · heatmap | accuracy | Improved |
| increase loss weights lambda1-3 to 3.0-4.0 (ablation) | DuoKD · 3D plot | accuracy | Improved |
Closest papers by meaning
- Large Language Model Meets Graph Neural Network in Knowledge Distillation (AAAI 2025 · similarity 0.92)
- VQGraph: Rethinking Graph Representation Space for Bridging GNNs and MLPs (ICLR 2024 · similarity 0.90)
- Preference-driven Knowledge Distillation for Few-shot Node Classification (NeurIPS 2025 · similarity 0.90)
- Bootstrapping Heterogeneous Graph Representation Learning via Large Language Models: A Generalized Approach (AAAI 2025 · similarity 0.89)
- LLM Enhancers for GNNs: An Analysis from the Perspective of Causal Mechanism Identification (ICML 2025 · similarity 0.89)
- UniGTE: Unified Graph–Text Encoding for Zero-Shot Generalization across Graph Tasks and Domains (NeurIPS 2025 · similarity 0.89)
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.
