Research map / Label-efficient and robust learning / Test-time adaptation
Neutralizing Token Aggregation via Information Augmentation for Efficient Test-Time Adaptation
Yizhe Xiong, Zihan Zhou, Yiwen Liang, Hui Chen, Zijia Lin, Xinhao Xu, Tianxiang Hao, Fan Zhang, Jungong Han, Guiguang Ding
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 [CLS] embedding augmentation (ablation) | NAVIA r=4 · four OOD benchmarks | avg accuracy (%): 66.3 -> 66.6 | Improved |
| add shallow layer [CLS] bias augmentation (ablation) | NAVIA r=4 · four OOD benchmarks | avg accuracy (%): 66.3 -> 66.8 | Improved |
| combine both augmentations (ablation) | NAVIA r=4 · four OOD benchmarks | avg accuracy (%): 66.3 -> 67.2 | Improved |
| add [CLS] embedding augmentation (ablation) | NAVIA r=8 · four OOD benchmarks | avg accuracy (%): 65.6 -> 66.2 | Improved |
| add shallow layer [CLS] bias augmentation (ablation) | NAVIA r=8 · four OOD benchmarks | avg accuracy (%): 65.6 -> 66.1 | Improved |
| combine both augmentations (ablation) | NAVIA r=8 · four OOD benchmarks | avg accuracy (%): 65.6 -> 66.7 | Improved |
| augment shallow instead of deep layers (ablation) | NAVIA · ImageNet-C | avg accuracy (%): 69.2 -> 69.8 | Improved |
| augment shallow instead of uniform layers (ablation) | NAVIA · ImageNet-C | avg accuracy (%): 69.2 -> 69.8 | Improved |
| tune pre-trained [CLS] instead of newly added tokens (ablation) | NAVIA · ImageNet-C | avg accuracy (%): 69.4 -> 69.8 | Improved |
| tune pre-trained [CLS] instead of re-initialized [CLS] (ablation) | NAVIA · ImageNet-C | avg accuracy (%): 65.9 -> 69.8 | Improved |
| increase token reduction from r=4 to r=8 | NAVIA · ImageNet-C | avg accuracy (%): 70.7 -> 69.8 | Hurt |
| use NAVIA_4 instead of ToMe_4 | ToMe · ImageNet-C | avg accuracy (%): 69.9 -> 70.7 | 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 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
- Efficient Test-time Adaptation through Candidate Verification and Divergence Shifts (NeurIPS 2026 · similarity 0.89)
- A Layer Selection Approach to Test Time Adaptation (AAAI 2025 · similarity 0.88)
- DOTA: Distributional Test-time Adaptation of Vision-Language Models (NeurIPS 2025 · similarity 0.88)
- SNAP: Low-Latency Test-Time Adaptation with Sparse Updates (NeurIPS 2025 · similarity 0.87)
- Backpropagation-Free Test-Time Adaptation via Probabilistic Gaussian Alignment (NeurIPS 2025 · similarity 0.87)
- SURGEON: Memory-Adaptive Fully Test-Time Adaptation via Dynamic Activation Sparsity (CVPR 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.
