Research map / Continual learning / Online and replay-based
Bi-Objective Continual Learning: Learning ‘New’ While Consolidating ‘Known’
Xiaoyu Tao, Xiaopeng Hong, Xinyuan Chang, Yihong Gong
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 pillar consolidation (PLC) loss with SOM pillars (ablation) | finetuning on new instances · CIFAR10 ST-NIL | accuracy (%): up to 9.22% over naive | Improved |
| add contrastive pillar (CPL) loss to PLC (ablation) | BOCL · CIFAR10 ST-NIL | accuracy (%): 76.98 -> 77.43 | Improved |
| use SOM pillars instead of random exemplars for CPL+PLC (ablation) | BOCL · CIFAR10 ST-NIL | accuracy (%): 76.62 -> 77.43 | Improved |
| add feature re-weighting omega to PLC with SOM pillars (ablation) | PLC · CIFAR10 ST-NIL | accuracy (%): 75.93 -> 76.98 | Improved |
| add feature re-weighting omega with random selection (ablation) | PLC · CIFAR10 ST-NIL | accuracy (%): 75.72 -> 75.72 | No clear effect |
| BOCL vs A-GEM* baseline on CIFAR100 full data | continual learning · CIFAR100 ST-NIL | accuracy (%): 51.24% -> 53.76% | Improved |
| BOCL vs IMM* on subImageNet full data | continual learning · subImageNet | top-5 accuracy (%): 48.52% -> 50.32% | Improved |
| finetune on hard or misclassified examples instead of random samples (ablation) | finetuning · CIFAR10 ST-NIL | accuracy (%) | Hurt |
| increase lambda from 0.1 to 10 (ablation) | BOCL · CIFAR10 ST-NIL | accuracy (%): 69.77 -> 72.10 | Improved |
| increase lambda from 10 to 50 (ablation) | BOCL · CIFAR10 ST-NIL | accuracy (%): 72.10 -> 72.06 | No clear effect |
| increase gamma from 0.1 to 10 (ablation) | BOCL · CIFAR10 ST-NIL | accuracy (%): 71.31 -> 72.10 | Improved |
| increase pillars from 10 to 225 (ablation) | BOCL · CIFAR10 ST-NIL | accuracy (%): 69.43 -> 72.10 | Improved |
Closest papers by meaning
- A continual learning survey: Defying forgetting in classification tasks (TPAMI 2021 · similarity 0.89)
- Train with Perturbation, Infer after Merging: A Two-Stage Framework for Continual Learning (NeurIPS 2025 · similarity 0.88)
- Pareto Continual Learning: Preference-Conditioned Learning and Adaption for Dynamic Stability-Plasticity Trade-off (AAAI 2025 · similarity 0.87)
- A Unified and General Framework for Continual Learning (ICLR 2024 · similarity 0.87)
- Do Your Best and Get Enough Rest for Continual Learning (CVPR 2025 · similarity 0.87)
- Mitigating Catastrophic Forgetting in Online Continual Learning by Modeling Previous Task Interrelations via Pareto Optimization (ICML 2024 · similarity 0.87)
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.
