Research map / Continual learning / Online and replay-based
Online Continual Learning with Contrastive Vision Transformer
Zhen Wang, Liu Liu, Yajing Kong, Jiaxian Guo, 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 focal contrastive loss (vs removing it) (ablation) | CVT · CIFAR100 Task-free, buffer 500 | accuracy (%): 19.92 -> 24.45 | Improved |
| add focal contrastive loss (vs removing it) (ablation) | CVT · CIFAR100 Task-free, buffer 500 | forgetting (%): 26.16 -> 21.86 | Improved |
| replace supervised contrastive loss with focal contrastive loss (ablation) | CVT · CIFAR100 Task-free, buffer 500 | accuracy (%): 20.73 -> 24.45 | Improved |
| replace focal contrastive loss with supervised contrastive loss (no forgetting gain over no contrastive loss) (ablation) | CVT · CIFAR100 Task-free, buffer 500 | forgetting (%): 26.16 -> 25.52 | No clear effect |
| add dual-classifier (vs removing it) (ablation) | CVT · CIFAR100 Task-free, buffer 500 | accuracy (%): 22.08 -> 24.45 | Improved |
| add dual-classifier (vs removing it) (ablation) | CVT · CIFAR100 Task-free, buffer 500 | forgetting (%): 29.81 -> 21.86 | Improved |
| add focal contrastive loss to ResNet18 (ablation) | ResNet18 · CIFAR100 Task-free, buffer 500 | accuracy (%): 15.72 -> 17.49 | Improved |
| add dual-classifier to ResNet18 (ablation) | ResNet18 · CIFAR100 Task-free, buffer 500 | accuracy (%): 15.72 -> 18.84 | Improved |
| replace ResNet18 with CVT architecture (minus L_FC) (ablation) | online CL · CIFAR100 Task-free, buffer 500 | accuracy (%): 15.72 -> 19.92 | Improved |
| add focal contrastive loss (vs removing it) (ablation) | CVT · TinyImageNet Task-free, buffer 500 | accuracy (%): 12.47 -> 14.71 | Improved |
| add dual-classifier (vs removing it) (ablation) | CVT · TinyImageNet Task-free, buffer 500 | accuracy (%): 13.63 -> 14.71 | Improved |
| replace focal contrastive loss with supervised contrastive loss (ablation) | CVT · TinyImageNet Task-free, buffer 500 | accuracy (%): 12.47 -> 12.59 | No clear effect |
Closest papers by meaning
- Provable Contrastive Continual Learning (ICML 2024 · similarity 0.89)
- Task-Free Dynamic Sparse Vision Transformer for Continual Learning (AAAI 2024 · similarity 0.88)
- Contrastive Continual Learning with Importance Sampling and Prototype-Instance Relation Distillation (AAAI 2024 · similarity 0.87)
- Federated Continual Learning Goes Online: Uncertainty-Aware Memory Management for Vision Tasks and Beyond (ICLR 2025 · similarity 0.87)
- Overcoming Catastrophic Forgetting in Visual Continual Learning with Reinforcement Fine-Tuning (NeurIPS 2026 · similarity 0.87)
- Enhancing Online Continual Learning with Plug-and-Play State Space Model and Class-Conditional Mixture of Discretization (CVPR 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.
