Research map / Label-efficient and robust learning / Test-time adaptation
Panda: Test-Time Adaptation with Negative Data Augmentation
Ruxi Deng, Wenxuan Bao, Tianxin Wei, Jingrui He
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 Panda negative-augmentation offset to Tent | Tent · CIFAR-10-C | accuracy: 62.8 -> 71.1 | Improved |
| add Panda to SAR | SAR · CIFAR-10-C | accuracy: 63.3 -> 70.7 | Improved |
| add Panda to DeYO | DeYO · CIFAR-100-C | accuracy: 38.0 -> 42.1 | Improved |
| add Panda to TTA methods generally | 10 TTA baselines · ImageNet-C | accuracy: +0.6 to +2.9 | Improved |
| add Panda to zero-shot CLIP | CLIP · CIFAR-10-C | accuracy: 59.0 -> 61.6 | Improved |
| add Panda to PDA method Zero | Zero · CIFAR-100-C | accuracy: 29.5 -> 33.3 | Improved |
| use Panda instead of positive augmentation (TPT, Zero, TPS) | TPS · CIFAR-10-C | accuracy: 63.7 -> 71.1 | Improved |
| use Panda offset instead of DeYO select-and-weight NDA | DeYO · CIFAR-10-C | accuracy: 65.5 -> 68.3 | Improved |
| share batch-wide negative augmentations vs shuffling patches within each image (ablation) | Panda · CIFAR-10-C | accuracy | Improved |
| average negative augmentation features before offset (ablation) | Panda · CIFAR-10-C | accuracy | Improved |
| reduce M/B ratio of negative augmentations (ablation) | Panda · CIFAR-10-C | accuracy | No clear effect |
| vary offset ratio beta (ablation) | Panda · CIFAR-10-C | accuracy | No clear effect |
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 3 of 3 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
- Efficient Test-time Adaptation through Candidate Verification and Divergence Shifts (NeurIPS 2026 · similarity 0.91)
- Noisy Test-Time Adaptation in Vision-Language Models (ICLR 2025 · similarity 0.91)
- Bayesian Test-Time Adaptation for Vision-Language Models (CVPR 2025 · similarity 0.90)
- Frustratingly Easy Test-Time Adaptation of Vision-Language Models (NeurIPS 2024 · similarity 0.90)
- The Illusion of Progress? A Critical Look at Test-Time Adaptation for Vision-Language Models (NeurIPS 2025 · similarity 0.90)
- Efficient Test-Time Adaptation of Vision-Language Models (CVPR 2024 · similarity 0.90)
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.
