Research map / Continual learning / Class-incremental
Asymmetric Synthetic Data Update for Domain Incremental Dataset Distillation
Minyoung Oh, Jae-Young Sim
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 asymmetric bi-level update to Finetune+Lstable (ablation) | Proposed · R-MNIST (IPC=10), domains 1-5 | accuracy: 26.9 -> 36.3 | Improved |
| add asymmetric bi-level update to Finetune+Lstable (ablation) | Proposed · R-MNIST (IPC=10), domains 16-20 | accuracy: 87.1 -> 94.5 | Improved |
| add asymmetric bi-level update to Finetune+Lstable (ablation) | Proposed · R-MNIST (IPC=20), domains 11-15 | accuracy: 45.0 -> 72.9 | Improved |
| add stability loss Lstable to Finetune (ablation) | Finetune · R-MNIST (IPC=10), domains 1-5 | accuracy: 33.4 -> 26.9 | Hurt |
| add stability loss Lstable to Finetune (ablation) | Finetune · R-MNIST (IPC=10), domains 11-15 | accuracy: 23.7 -> 59.3 | Improved |
| add stability loss Lstable to Finetune (ablation) | Finetune · R-MNIST (IPC=20), domains 16-20 | accuracy: 85.6 -> 77.5 | Hurt |
| use linear alpha/beta assignment instead of bi-level optimization (ablation) | Proposed · R-MNIST (IPC=10), domains 1-5 | accuracy: 36.3 -> 31.0 | Hurt |
| use bi-level learned update rates instead of linear assignment (ablation) | Proposed · R-MNIST (IPC=20), domains 1-5 | accuracy: 27.3 -> 32.9 | Improved |
| reduce distillation iterations to 1,000 (ablation) | Proposed · PACS (IPC=10) | accuracy: 48.0 -> 46.3 | Mixed |
| add proposed method over Finetune | Proposed · PACS (IPC=10) | average accuracy: 26.9 -> 48.0 | Improved |
| add proposed method over MAS | Proposed · R-MNIST (IPC=10) | average accuracy: 45.6 -> 62.7 | Improved |
| add proposed method over Finetune | Proposed · Seq-CORe50 (IPC=20) | forgetting: 79.2 -> 38.7 | Improved |
Closest papers by meaning
- Large Scale Dataset Distillation with Domain Shift (ICML 2024 · similarity 0.89)
- Embarrassingly Simple Dataset Distillation (ICLR 2024 · similarity 0.89)
- Data Distillation Can Be Like Vodka: Distilling More Times For Better Quality (ICLR 2024 · similarity 0.89)
- Towards Lossless Dataset Distillation via Difficulty-Aligned Trajectory Matching (ICLR 2024 · similarity 0.88)
- Rethinking Dataset Distillation for Classification: Do Distilled Sets Outperform Coresets? (NeurIPS 2026 · similarity 0.88)
- Group Distributionally Robust Dataset Distillation with Risk Minimization (ICLR 2025 · 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.
