Research map / Graph learning / Graph self-supervised learning
HiLoMix: Robust High- and Low-Frequency Graph Learning Framework for Mixing Address Association
Xiaofan Tu, Tiantian Duan, Shuyi Miao, Hanwen Zhang, Yi Sun
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 heterogeneous base models (LR, RF, MLP) to stacking (ablation) | HiLoMix · Mixing address association (Tornado Cash) | F1: 0.7393 -> 0.8382 | Improved |
| add stacking framework over high/low GNNs (ablation) | HiLoMix · Mixing address association (Tornado Cash) | F1: 0.7882 -> 0.8382 | Improved |
| add high/low-pass GNNs to the ensemble (ablation) | HiLoMix · Mixing address association (Tornado Cash) | F1: 0.8075 -> 0.8382 | Improved |
| replace GCNs with high/low-pass GNNs (ablation) | HiLoMix · Mixing address association (Tornado Cash) | MRR: 0.4228 -> 0.4612 | Improved |
| add mixing transaction edges (heterogeneous graph) (ablation) | HiLoMix · Mixing address association (Tornado Cash) | AUC: 0.8931 -> 0.9137 | Improved |
| add confidence-based label division (ablation) | HiLoMix · Mixing address association (Tornado Cash) | F1: 0.8238 -> 0.8382 | Improved |
| add high-pass GNN (ablation) | HiLoMix · Mixing address association (Tornado Cash) | F1: 0.8169 -> 0.8382 | Improved |
| add low-pass GNN (ablation) | HiLoMix · Mixing address association (Tornado Cash) | F1: 0.8286 -> 0.8382 | Improved |
| use 197-dim node features instead of 221-dim (ablation) | HiLoMix · Mixing address association (Tornado Cash) (feature sets) | F1: 0.8183 -> 0.8382 | Improved |
| use 197-dim features instead of MixBroker features (ablation) | HiLoMix · Mixing address association (Tornado Cash) (feature sets) | F1: 0.8248 -> 0.8382 | Improved |
| use HiLoMix instead of best baseline NRGL | NRGL · Mixing address association (Tornado Cash) | AUC: 0.8513 -> 0.9137 | Improved |
| 5-fold vs other folds in stacking CV (ablation) | HiLoMix · Mixing address association (Tornado Cash) | F1: 83.8 vs 80.9-83.0 | No clear effect |
Closest papers by meaning
- Dynamic Neighborhood Modeling via Node-Subgraph Contrastive Learning for Graph-Based Fraud Detection (AAAI 2025 · similarity 0.84)
- Clean-Label Graph Backdoor Attack in the Node Classification Task (AAAI 2025 · similarity 0.83)
- Does GCL Need a Large Number of Negative Samples? Enhancing Graph Contrastive Learning with Effective and Efficient Negative Sampling (AAAI 2025 · similarity 0.83)
- Partitioning Message Passing for Graph Fraud Detection (ICLR 2024 · similarity 0.83)
- Hypergraph Representation Learning with Hyperlink Random Effects (NeurIPS 2026 · similarity 0.83)
- TopoGCL: Topological Graph Contrastive Learning (AAAI 2024 · similarity 0.83)
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.
