Research map / Label-efficient and robust learning / Self-supervised and contrastive
Co2L: Contrastive Continual Learning
Hyuntak Cha, Jaeho Lee, Jinwoo Shin
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 IRD (with replay buffer) (ablation) | Co2L · Seq-CIFAR-10 Class-IL | accuracy (%): 53.57 -> 65.57 | Improved |
| add IRD (without replay buffer) (ablation) | Co2L · Seq-CIFAR-10 Class-IL | accuracy (%): 53.25 -> 58.89 | Improved |
| add buffered samples to representation learning (with IRD) (ablation) | Co2L · Seq-CIFAR-10 Class-IL | accuracy (%): 58.89 -> 65.57 | Improved |
| add buffered samples without IRD (ablation) | Co2L · Seq-CIFAR-10 Class-IL | accuracy (%): 53.25 -> 53.57 | No clear effect |
| use asymmetric SupCon instead of original SupCon (ablation) | Co2L · Seq-CIFAR-10 Class-IL (buffer 200) | accuracy (%): 60.49 -> 65.57 | Improved |
| use asymmetric SupCon instead of original SupCon (ablation) | Co2L · Seq-CIFAR-10 Class-IL (buffer 500) | accuracy (%): 68.66 -> 74.26 | Improved |
| use asymmetric SupCon instead of original SupCon (ablation) | Co2L · Seq-Tiny-ImageNet Class-IL (buffer 200) | accuracy (%): 13.51 -> 13.88 | Improved |
| use asymmetric SupCon instead of original SupCon (ablation) | Co2L · Seq-Tiny-ImageNet Class-IL (buffer 500) | accuracy (%): 19.68 -> 20.12 | Improved |
| use IRD instead of SEED embedding distillation (ablation) | Co2L · Seq-CIFAR-10 Class-IL (buffer 200) | accuracy (%): 53.42 -> 65.57 | Improved |
| use IRD instead of MSE embedding matching (ablation) | Co2L · Seq-CIFAR-10 Class-IL (buffer 500) | accuracy (%): 62.83 -> 74.26 | Improved |
| use IRD instead of MSE projection matching (ablation) | Co2L · Seq-CIFAR-10 Class-IL (buffer 500) | accuracy (%): 57.47 -> 74.26 | Improved |
| use IRD instead of MSE projection matching (ablation) | Co2L · Seq-Tiny-ImageNet Task-IL (buffer 200) | accuracy (%): 34.38 -> 42.37 | Improved |
Closest papers by meaning
- Self-Supervised Models Are Continual Learners (CVPR 2022 · similarity 0.90)
- Contrastive Continual Learning with Importance Sampling and Prototype-Instance Relation Distillation (AAAI 2024 · similarity 0.90)
- Provable Contrastive Continual Learning (ICML 2024 · similarity 0.90)
- Probing Representation Forgetting in Supervised and Unsupervised Continual Learning (CVPR 2022 · similarity 0.89)
- Enhancing Visual Continual Learning with Language-Guided Supervision (CVPR 2024 · similarity 0.88)
- Measuring Representational Shifts in Continual Learning: A Linear Transformation Perspective (ICML 2025 · 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.
