Research map / Continual learning / Domain and prompt-based
Reshaping the Online Data Buffering and Organizing Mechanism for Continual Test-Time Adaptation
Zhilin Zhu, Xiaopeng Hong, Zhiheng Ma, Weijun Zhuang, YaoHui Ma, Yong Dai, Yaowei Wang
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 |
|---|---|---|---|
| replace reservoir buffering with uncertainty-aware buffering (with L_PCE) (ablation) | OBAO · CIFAR100-to-CIFAR100C | mean error (%): 30.5 -> 29.7 | Improved |
| use FIFO buffering instead of uncertainty-aware buffering (ablation) | OBAO · CIFAR100-to-CIFAR100C | mean error (%): 30.7 -> 30.6 | No clear effect |
| add class relation preservation constraint L_CRP alone (ablation) | self-training baseline · CIFAR100-to-CIFAR100C | mean error (%): 30.7 -> 30.0 | Improved |
| add L_CRP on top of uncertainty buffering with replay (ablation) | OBAO · CIFAR100-to-CIFAR100C | mean error (%): 29.7 -> 29.0 | Improved |
| add uncertainty-aware buffer with pseudo-target replay (ablation) | self-training baseline · CIFAR100-to-CIFAR100C | mean error (%): 30.7 -> 29.7 | Improved |
| increase lambda_CRP over a broad range, but set 300 (ablation) | OBAO · CIFAR100-to-CIFAR100C | mean error (%) | Mixed |
| use OBAO instead of CoTTA on long-horizon segmentation | Segformer-B5 · Cityscapes-to-ACDC (10 rounds) | mean mIoU (%): 58.6 -> 61.3 | Improved |
| use entropy-minimizing TENT-cont. over long sequences | Segformer-B5 · Cityscapes-to-ACDC (10 rounds) | mIoU (%): 56.7 -> 47.8 by round 10 | Hurt |
| use OBAO instead of DSS on ImageNet-C | ResNet-50 · ImageNet-to-ImageNet-C | mean error (%): 64.6 -> 59.0 | Improved |
| TENT-cont. on CIFAR100-to-CIFAR100C long sequence | ResNeXt-29 · CIFAR100-to-CIFAR100C | mean error (%): 35.4 (BN adapt) -> 60.9 | Hurt |
Closest papers by meaning
- A Versatile Framework for Continual Test-Time Domain Adaptation: Balancing Discriminability and Generalizability (CVPR 2024 · similarity 0.89)
- Continual-MAE: Adaptive Distribution Masked Autoencoders for Continual Test-Time Adaptation (CVPR 2024 · similarity 0.88)
- Class-aware Domain Knowledge Fusion and Fission for Continual Test-Time Adaptation (NeurIPS 2025 · similarity 0.88)
- DPCore: Dynamic Prompt Coreset for Continual Test-Time Adaptation (ICML 2025 · similarity 0.88)
- Buffer layers for Test-Time Adaptation (NeurIPS 2025 · similarity 0.88)
- ReservoirTTA: Prolonged Test-time Adaptation for Evolving and Recurring Domains (NeurIPS 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.
