Research map / Continual learning / Class-incremental
Subspace-Aware Feature Reshaping for Open-Set Graph Class-Incremental Learning
Weichao Zhang, Shuai Zheng, Yeyu Yan, Zhizhe Liu, Zhenfeng Zhu, Yao Zhao
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 energy-based task routing (ablation) | SAFER · CoraFull | OSCR: 4.4 -> 94.6 | Improved |
| add energy-based task routing (ablation) | SAFER · Products | OSCR: 6.8 -> 97.1 | Improved |
| add task-specific classifier heads instead of a single head (ablation) | SAFER · CoraFull | OSCR: 90.49 -> 94.6 | Improved |
| add task-specific classifier heads instead of a single head (ablation) | SAFER · Products | AF: -13.6 -> 0.0 | Improved |
| add node-adaptive prompts (ablation) | SAFER · CoraFull | OSCR: 88.1 -> 94.6 | Improved |
| add node-adaptive prompts (ablation) | SAFER · Products | OSCR: 95.8 -> 97.1 | Improved |
| add PPR diffusion (ablation) | SAFER · CoraFull | OSCR: 93.3 -> 94.6 | Improved |
| add PPR diffusion (ablation) | SAFER · CoraFull | AP: 94.8 -> 94.9 | No clear effect |
| add PPR diffusion (ablation) | SAFER · Products | OSCR: 95.8 -> 97.1 | Improved |
| add prompts and PPR together (ablation) | SAFER · CoraFull | AP: 88.9 -> 94.9 | Improved |
| use subspace energy score instead of max softmax for unknown detection (ablation) | SAFER · Reddit | AUROC: 75.6% -> 100% | Improved |
| SAFER vs TPP with 5 classes per task | TPP · Arxiv (5 classes/task) | OSCR: 58.5 -> 81.8 | Improved |
Closest papers by meaning
- Replay-and-Forget-Free Graph Class-Incremental Learning: A Task Profiling and Prompting Approach (NeurIPS 2024 · similarity 0.91)
- What Matters in Graph Class Incremental Learning? An Information Preservation Perspective (NeurIPS 2024 · similarity 0.89)
- GACL: Exemplar-Free Generalized Analytic Continual Learning (NeurIPS 2024 · similarity 0.87)
- GraphKeeper: Graph Domain-Incremental Learning via Knowledge Disentanglement and Preservation (NeurIPS 2025 · similarity 0.87)
- Order-Robust Class Incremental Learning: Graph-Driven Dynamic Similarity Grouping (CVPR 2025 · similarity 0.87)
- Navigating Semantic Drift in Task-Agnostic Class-Incremental Learning (ICML 2025 · 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.
