Research map / Continual learning / Domain and prompt-based
Online Versatile Incremental Learning: Towards Class and Domain-Agnostic Adaptation at Any Time
Jaeho Lee, Jun-Yeong Moon, Min-Yeong Park, Jung Uk Kim, Gyeong-Moon Park
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 DFM (ablation) | MVP baseline · Online VIL CORe50 (no replay, MVP base) | A_Last: 52.84 -> 64.57 | Improved |
| add GTP (ablation) | MVP baseline · Online VIL CORe50 (no replay, MVP base) | A_Last: 52.84 -> 65.28 | Improved |
| add GTP on top of DFM (ablation) | TopFlow · Online VIL CORe50 (no replay, MVP base) | A_Last: 64.57 -> 66.20 | Improved |
| apply DFM at early layer pair (0,5) (ablation) | DFM loss · DFM layer pair ablation (CORe50, w/o GTP) | A_Last: 52.84 -> 49.92 | Hurt |
| apply DFM at layer pair (6,11) instead of (5,10) (ablation) | DFM loss · DFM layer pair ablation (CORe50, w/o GTP) | A_Last: 52.98 -> 54.82 | Improved |
| add DFM and GTP to CODA-P (ablation) | CODA-P · Online VIL CORe50 (CODA-P base) | A_Last: 49.13 -> 54.62 | Improved |
| add DFM and GTP to PEC (ablation) | PEC · Online VIL CORe50 (PEC base) | A_Last: 45.99 -> 48.86 | Improved |
| add DFM only to CODA-P in standard CIL (ablation) | CODA-P · CIFAR-100 10-split CIL (CODA-P) | accuracy: 84.17 -> 84.19 | No clear effect |
| add DFM and GTP to S-Prompt in DIL (ablation) | S-Prompt · CORe50 8-split DIL (S-Prompt) | accuracy: 82.96 -> 87.50 | Improved |
| use TopFlow instead of MVP (no replay) | MVP · Online VIL CORe50 (no replay) | A_Last: 52.84 -> 66.20 | Improved |
| add replay buffer 500 to TopFlow on CORe50 | TopFlow · Online VIL CORe50 | A_Last: 66.20 -> 91.14 | Improved |
| use TopFlow instead of ICON in Offline VIL (CORe50) | ICON · Offline VIL CORe50 | Avg. Acc: 79.61 -> 79.77 | No clear effect |
Closest papers by meaning
- CTP:Towards Vision-Language Continual Pretraining via Compatible Momentum Contrast and Topology Preservation (ICCV 2023 · similarity 0.88)
- Class-Domain Incremental Learning with Extensible Multi-Center Modeling (NeurIPS 2026 · similarity 0.88)
- Topology-aware Knowledge Preservation for Class-Incremental Learning (AAAI 2026 · similarity 0.87)
- Advancing Cross-domain Discriminability in Continual Learning of Vision-Language Models (NeurIPS 2024 · similarity 0.86)
- Enhancing Online Continual Learning with Plug-and-Play State Space Model and Class-Conditional Mixture of Discretization (CVPR 2025 · similarity 0.86)
- Contextual Flow Matching: Adaptive Step Selection in Flow Models for Efficient Visual Generation (NeurIPS 2026 · similarity 0.85)
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.
