Research map / Label-efficient and robust learning / Test-time adaptation
When and Where to Reset Matters for Long-Term Test-Time Adaptation
Taejun Lim, Joong-Won Hwang, Kibok Lee
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 adaptive (collapse-triggered) reset instead of fixed-interval reset (ablation) | ASR on ROID · CCC mean | accuracy (%): 35.86 -> 38.32 | Improved |
| add selective (partial) reset instead of full-parameter reset (ablation) | ASR on ROID · CCC mean | accuracy (%): 35.39 -> 38.32 | Improved |
| add importance-aware regularizer term F-bar (ablation) | ASR on ROID · CCC mean | accuracy (%): 37.80 -> 38.32 | Improved |
| add adaptive lambda0 (ablation) | ASR on ROID · CCC mean | accuracy (%): 37.89 -> 38.32 | Improved |
| add adaptive mu0 (ablation) | ASR on ROID · CCC mean | accuracy (%): 37.79 -> 38.32 | Improved |
| add all five ASR components to ROID (ablation) | ROID · CCC mean | accuracy (%): 33.91 -> 38.32 | Improved |
| add ASR reset to ROID instead of RDumb | ROID · CCC-Hard (ResNet-50) | accuracy (%): 15.41 -> 22.21 | Improved |
| add ASR to ETA | ETA · CCC mean (ResNet-50) | accuracy (%): 20.86 -> 36.73 | Improved |
| add ASR to EATA | EATA · CCC mean (ResNet-50) | accuracy (%): 29.84 -> 35.80 | Improved |
| add COME to ROID | ROID · CCC mean (ResNet-50) | accuracy (%): 34.28 -> 0.12 | Hurt |
| add RDumb periodic reset to ROID under modified CCC-Hard (random durations) | ROID · CCC-Hard modified (seed 43) | accuracy (%): 16.51 -> 15.99 | Hurt |
| add ASR to ROID under modified CCC-Hard (random durations) | ROID · CCC-Hard modified (seed 43) | accuracy (%): 16.51 -> 21.56 | Improved |
Closest papers by meaning
- Persistent Test-time Adaptation in Recurring Testing Scenarios (NeurIPS 2024 · similarity 0.90)
- RDumb: A simple approach that questions our progress in continual test-time adaptation (NeurIPS 2023 · similarity 0.89)
- A Probabilistic Framework for Lifelong Test-Time Adaptation (CVPR 2023 · similarity 0.89)
- ReservoirTTA: Prolonged Test-time Adaptation for Evolving and Recurring Domains (NeurIPS 2025 · similarity 0.87)
- A Layer Selection Approach to Test Time Adaptation (AAAI 2025 · similarity 0.87)
- DPCore: Dynamic Prompt Coreset for Continual Test-Time Adaptation (ICML 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.
