Research map / Architectures and efficiency / Pruning and efficient tuning
Towards Robust Edge Model Adaptation via Elastic Architecture Search
Xianhang Chu, Xu Yang, Kun Wei, Xi 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 |
|---|---|---|---|
| add channel-width monotonicity constraint to search space (ablation) | Elastic search · ImageNet (6M budget) | est. top-1 %: 57.4 -> 59.6 | Improved |
| add channel-width monotonicity constraint to search space (ablation) | Elastic search · ImageNet (15M budget) | est. top-1 %: 63.0 -> 65.8 | Improved |
| add low-rank online fine-tuning to searched model (ablation) | EEAS · ImageNet-1K (400 MB) | top-1 %: 73.1 -> 75.3 | Improved |
| add knowledge reflux and re-search (ablation) | EEAS · ImageNet-1K (400 MB) | top-1 %: 75.3 -> 76.5 | Improved |
| add knowledge reflux and re-search (ablation) | EEAS · ImageNet-1K (NAS comparison) | top-1 %: 76.5 -> 77.9 | Improved |
| increase simulation sampling number K from 1 toward 3 (ablation) | MCTS search · ImageNet | top-1 % | Improved |
| add SAR / REM / PALM auxiliary losses to CE adaptation (ablation) | Ours (CE) · ImageNet-C | avg accuracy %: 56.23 -> 58.01 (with PALM) | Improved |
| use binary elastic searched model instead of ViT-S baseline | CTTA · CIFAR100-C | avg accuracy %: 74.6 (PALM ViT-S) -> 76.9 (Ours+PALM) | Improved |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 2 of 2 on most shared results.
- HEP-NAS — HEP-NAS: Towards Efficient Few-shot Neural Architecture Search via Hierarchical Edge Partitioning (AAAI 2025) · better on 2 of 2 shared results
- IS-DARTS — IS-DARTS: Stabilizing DARTS through Precise Measurement on Candidate Importance (AAAI 2024) · better on 1 of 1 shared result
Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.
Closest papers by meaning
- SNAP: Low-Latency Test-Time Adaptation with Sparse Updates (NeurIPS 2025 · similarity 0.88)
- AdMiT: Adaptive Multi-Source Tuning in Dynamic Environments (CVPR 2025 · similarity 0.88)
- Towards Robust and Efficient Cloud-Edge Elastic Model Adaptation via Selective Entropy Distillation (ICLR 2024 · similarity 0.88)
- ViDA: Homeostatic Visual Domain Adapter for Continual Test Time Adaptation (ICLR 2024 · similarity 0.85)
- PALM: Pushing Adaptive Learning Rate Mechanisms for Continual Test-Time Adaptation (AAAI 2025 · similarity 0.85)
- DPCore: Dynamic Prompt Coreset for Continual Test-Time Adaptation (ICML 2025 · 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.
