Research map / Federated learning and privacy / Personalized and heterogeneous FL
Towards Stable Federated Continual Test-Time Adaptation in Wild World
Liwen Wang, Xingbo Dong, Iman Yi Liao, Zhe Jin
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 BPA (MAP-regularized local adaptation) instead of Tent (ablation) | FedAvg global model · CIFAR10-C | mean accuracy (%): 64.72 -> 67.15 | Improved |
| add BPA instead of Tent (ablation) | FedAvg global model · CIFAR100-C | mean accuracy (%): 57.51 -> 62.89 | Improved |
| add UGSA aggregation to Tent (without BPA) (ablation) | Tent · Cityscapes-to-ACDC Seq4 | mean mIoU (%): 52.72 -> 58.12 | Improved |
| add UGSA aggregation to Tent (without BPA) (ablation) | Tent · CIFAR10-C | mean accuracy (%): 64.72 -> 66.83 | Improved |
| combine BPA with UGSA (full method) vs BPA alone (ablation) | BPFedCTTA · Cityscapes-to-ACDC Seq4 | mean mIoU (%): 58.30 -> 65.19 | Improved |
| combine BPA with UGSA (full method) vs BPA alone (ablation) | BPFedCTTA · CIFAR100-C (ablation table) | mean accuracy (%): 62.89 -> 64.57 | Improved |
| switch UGSA gating off (beta = 0.0) (ablation) | BPFedCTTA · Cityscapes-to-ACDC | Stability (%): 62.9 -> 51.8 | Hurt |
| lower server learning rate gamma to 0.1 (ablation) | UGSA · Cityscapes-to-ACDC | Plasticity (%): 63.5 mIoU (gamma 0.5) -> 61.0 plasticity | Hurt |
| raise server learning rate gamma to 1.0 (ablation) | UGSA · Cityscapes-to-ACDC | Stability (%): -> 59.0 | Hurt |
| replace predictive entropy with energy score for uncertainty in UGSA (ablation) | UGSA · CIFAR10-C | mean accuracy (%): 68.44 -> 68.62 | Improved |
| replace predictive entropy with MC Dropout (T=10) or ensemble for uncertainty (ablation) | UGSA · CIFAR10-C | time per client (s): 4.492 -> 64.920 (MC Dropout); 4.492 -> 20.125 (Ensemble) | Mixed |
| replace predictive entropy with max probability for uncertainty (ablation) | UGSA · CIFAR10-C | mean accuracy (%): 68.44 -> 68.10 | Hurt |
Closest papers by meaning
- TTA-FedDG: Leveraging Test-Time Adaptation to Address Federated Domain Generalization (AAAI 2025 · similarity 0.88)
- Dual-Personalizing Adapter for Federated Foundation Models (NeurIPS 2024 · similarity 0.87)
- Enhancing Federated Domain Adaptation with Multi-Domain Prototype-Based Federated Fine-Tuning (ICLR 2025 · similarity 0.87)
- HERO: A Heterogeneity-Aware Benchmark Library for Federated Continual Learning (NeurIPS 2026 · similarity 0.87)
- FLUX: Efficient Descriptor-Driven Clustered Federated Learning under Arbitrary Distribution Shifts (NeurIPS 2025 · similarity 0.86)
- Federated Learning with Sample-level Client Drift Mitigation (AAAI 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.
