Research map / Graph learning / Graph self-supervised learning
DGP: A Dual-Granularity Prompting Framework for Fraud Detection with Graph-Enhanced LLMs
Yuan Li, Jun Hu, Bryan Hooi, Bingsheng He, Cheng Chen
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 |
|---|---|---|---|
| summarize neighbors into coarse-grained text while keeping the target node fine-grained (DGP) instead of concatenating neighbor texts | Qwen3-8B graph-enhanced LLM · YelpReviews | AUPRC (%): 38.36 -> 48.87 (vs InstructGLM) | Improved |
| add graph context and summarization to a standalone LLM | Qwen3-8B · YelpReviews | AUPRC (%): 30.45 -> 48.87 | Improved |
| add textual summarization (ablation) | DGP · YelpReviews and AmazonVideo | MacroF1, AUROC, AUPRC | Improved |
| add numerical summarization (mean aggregation) (ablation) | DGP · YelpReviews and AmazonVideo | MacroF1, AUROC, AUPRC | Improved |
| add Markov Diffusion Kernel metapath trimming (ablation) | DGP · YelpReviews and AmazonVideo | MacroF1, AUROC, AUPRC | Improved |
| add metapath-level summarization (ablation) | DGP · YelpReviews and AmazonVideo | MacroF1, AUROC, AUPRC | Improved |
| use task-agnostic instead of task-aware summarization prompts (ablation) | DGP · YelpReviews | AUPRC (%): 29.02 -> 48.87 | Improved |
| use task-agnostic instead of task-aware summarization prompts (ablation) | DGP · AmazonVideo | AUPRC (%): 31.82 -> 34.63 | Improved |
| shrink summarization budget to 5 tokens (ablation) | DGP · YelpReviews and AmazonVideo | MacroF1, AUROC, AUPRC | Hurt |
| lengthen summarization budget to 80 tokens (ablation) | DGP · YelpReviews and AmazonVideo | MacroF1, AUROC, AUPRC | Hurt |
| use GNNs instead of MLP on heterogeneous fraud graphs | baselines · YelpReviews | AUPRC (%): 32.24 -> 42.11 (MLP vs ConsisGAD) | Improved |
| encoding-based prompting (GraphGPT, HiGPT) instead of text-only dual granularity prompting | graph-enhanced LLMs · YelpReviews | AUPRC (%): 30.66 (GraphGPT) vs 48.87 (DGP) | Hurt |
Closest papers by meaning
- Dynamic Neighborhood Modeling via Node-Subgraph Contrastive Learning for Graph-Based Fraud Detection (AAAI 2025 · similarity 0.88)
- DiG-In-GNN: Discriminative Feature Guided GNN-Based Fraud Detector against Inconsistencies in Multi-Relation Fraud Graph (AAAI 2024 · similarity 0.88)
- Context-aware Graph Neural Network for Graph-based Fraud Detection with Extremely Limited Labels (AAAI 2025 · similarity 0.88)
- A Label-free Heterophily-guided Approach for Unsupervised Graph Fraud Detection (AAAI 2025 · similarity 0.87)
- Partitioning Message Passing for Graph Fraud Detection (ICLR 2024 · similarity 0.86)
- Barely Supervised Learning for Graph-Based Fraud Detection (AAAI 2024 · similarity 0.86)
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.
