Research map / Label-efficient and robust learning / Test-time adaptation
Curvature-Aware Zeroth-Order Optimization for Memory-Efficient Test-Time Adaptation
Junming Zhang, Shuyu Yin, Peilin Liu, Rendong Ying, Fei Wen
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 curvature-aware anisotropic perturbation instead of isotropic Gaussian (ablation) | ZO (k=20) · ImageNet-C sev5 (reset per corruption) | average accuracy (%): 62.9 -> 69.0 | Improved |
| increase number of perturbations k from 2 to 20 (ablation) | CAZO · ImageNet-C sev5 (reset per corruption) | average accuracy (%): 65.2 -> 69.0 | Improved |
| increase k from 2 to 6 (ablation) | CAZO · ImageNet-C Gaussian noise sev5 | accuracy (%): 59.3 -> 62.1 | Improved |
| increase k beyond 6 (ablation) | CAZO · ImageNet-C Gaussian noise sev5 | accuracy and ECE | No clear effect |
| use single-point perturbation instead of symmetric multi-point (ablation) | CAZO · ImageNet-C | accuracy | Hurt |
| EMA coefficient nu 0.8 versus 1.0 (ablation) | CAZO · ImageNet-C Gaussian noise sev5 | accuracy: 69.0 | Hurt |
| place adapter at layer 3 instead of later layers (ablation) | CAZO · ImageNet-C Gaussian noise sev5 | accuracy | Improved |
| use smaller adapter down-sampling ratios (96, 128) with more parameters (ablation) | CAZO · ImageNet-C Gaussian noise sev5 | accuracy | No clear effect |
| swap BP-based TENT for CAZO forward-only adaptation | TENT · ImageNet-C sev5 | memory (MB): 6,404 -> 1,695 | Improved |
| use CAZO instead of ZO under 8-bit quantization | ZO · ImageNet-C sev5 (8-bit) | average accuracy (%): 61.8 -> 67.8 | Improved |
| use CAZO instead of ZO under 6-bit quantization | ZO · ImageNet-C sev5 (6-bit) | average accuracy (%): 55.6 -> 61.2 | Improved |
| use CAZO instead of FOA | FOA · ImageNet-C sev5 (reset per corruption) | average accuracy (%): 65.8 -> 69.0 | Improved |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 1 on most shared results.
- TENT — Tent: Fully Test-Time Adaptation by Entropy Minimization (ICLR 2021) · better on 2 of 2 shared results
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
- Refining Adaptive Zeroth-Order Optimization at Ease (ICML 2025 · similarity 0.88)
- Test-Time Adaptation on Noisy Data via Model-Pruning-Based Filtering and Flatness-Aware Entropy Minimization (AAAI 2025 · similarity 0.88)
- A Layer Selection Approach to Test Time Adaptation (AAAI 2025 · similarity 0.88)
- SNAP: Low-Latency Test-Time Adaptation with Sparse Updates (NeurIPS 2025 · similarity 0.87)
- EVA-0: Test-Time Model Evolution with Only Two Forward Passes per Sample (NeurIPS 2026 · similarity 0.87)
- Backpropagation-Free Test-Time Adaptation via Probabilistic Gaussian Alignment (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.
