Research map / Continual learning / Online and replay-based
CGLB: Benchmark Tasks for Continual Graph Learning
Xikun ZHANG, Dongjin Song, Dacheng Tao
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 MAS regularization to bare GCN | Bare GCN · CoraFull-CL task-IL | AP (%): 58.0±1.7 -> 93.0±0.5 | Improved |
| add EWC regularization to bare GCN | Bare GCN · CoraFull-CL task-IL | AP (%): 58.0±1.7 -> 78.9±2.4 | Improved |
| add LwF distillation to bare GCN | Bare GCN · CoraFull-CL task-IL | AP (%): 58.0±1.7 -> 56.1±2.0 | No clear effect |
| add LwF distillation to bare GCN | Bare GCN · Arxiv-CL task-IL | AP (%): 61.7±3.8 -> 84.2±0.5 | Improved |
| keep inter-task edges | EWC · Products-CL task-IL | AP (%): 90.1±0.3 -> 73.6±1.3 | Hurt |
| keep inter-task edges | MAS · Products-CL task-IL | AP (%): 91.2±0.6 -> 78.3±1.6 | Hurt |
| keep inter-task edges | Bare model · Reddit-CL task-IL | AP (%): 73.6±3.5 -> 67.4±5.5 | Hurt |
| keep inter-task edges | EWC · Arxiv-CL task-IL | AP (%): 78.8±2.7 -> 67.6±3.2 | Hurt |
| keep inter-task edges | Joint training · Products-CL task-IL | AP (%): 91.8±0.2 -> 87.6±0.5 | Hurt |
| keep inter-task edges | TWP · Products-CL task-IL | AP (%): 90.3±0.1 -> 88.2±0.8 | Hurt |
| switch from task-IL to class-IL | Bare model · CoraFull-CL | AP (%): 58.0±1.7 -> 2.9±0.0 | Hurt |
| add ER-GNN memory replay to bare GCN | Bare GCN · Reddit-CL class-IL | AP (%): 5.1±0.3 -> 88.5±2.3 | Improved |
Closest papers by meaning
- A Selective Learning Method for Temporal Graph Continual Learning (ICML 2025 · similarity 0.90)
- Exploring Rationale Learning for Continual Graph Learning (AAAI 2025 · similarity 0.88)
- Efficient Graph Continual Learning via Lightweight Graph Neural Tangent Kernels-based Dataset Distillation (ICML 2025 · similarity 0.88)
- No Metric to Rule Them All: Toward Principled Evaluations of Graph-Learning Datasets (ICML 2025 · similarity 0.88)
- Replay-and-Forget-Free Graph Class-Incremental Learning: A Task Profiling and Prompting Approach (NeurIPS 2024 · similarity 0.88)
- Dynamic Sub-graph Distillation for Robust Semi-supervised Continual Learning (AAAI 2024 · similarity 0.88)
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.
