Research map / Label-efficient and robust learning / Test-time adaptation
On the Robustness of Open-World Test-Time Training: Self-Training with Dynamic Prototype Expansion
Yushu Li, Xun Xu, Yongyi Su, Kui Jia
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 distribution alignment on top of prototype clustering and expansion (ablation) | OWTTT · CIFAR10-C + noise | Acc_H: 81.4 -> 91.6 | Improved |
| add prototype expansion to prototype clustering (ablation) | OWTTT · CIFAR10-C + SVHN | Acc_H: 72.7 -> 77.4 | Improved |
| add plain prototype clustering with source prototypes only (ablation) | OWTTT with O.D. · CIFAR10-C + noise | Acc_H: 81.4 -> 76.1 | Hurt |
| enable strong OOD detector at inference (ablation) | TEST baseline · CIFAR10-C + noise | Acc_N: 0.0 -> 100.0 | Improved |
| use adaptive threshold tau* instead of fixed thresholds 0.1 to 0.9 (ablation) | strong OOD pruning · CIFAR10-C with strong OOD | strong OOD detection | Improved |
| use distribution alignment (BN, TTAC) instead of self-training (TENT, SHOT) | baselines under strong OOD · CIFAR10-C | Acc_H: BN 85.11 vs TENT 32.77 (noise) | Improved |
| vary ratio of strong to weak OOD from 0.2 to 1.0 (ablation) | OWTTT · CIFAR10-C | Acc_H: noise 90.83 -> 91.34 | No clear effect |
| enable OOD detector when target has no strong OOD (ablation) | OWTTT · clean CIFAR10-C | Acc_S: 85.25 -> 78.63 | Hurt |
| OWTTT vs best baselines on ImageNet-C noise | OWTTT · ImageNet-C + noise | Acc_H: SHOT 63.75 vs Ours 58.56 | Hurt |
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
- Towards Real-World Test-Time Adaptation: Tri-net Self-Training with Balanced Normalization (AAAI 2024 · similarity 0.87)
- IT$^3$: Idempotent Test-Time Training (ICML 2025 · similarity 0.86)
- Rethinking the Evaluation Protocol of Domain Generalization (CVPR 2024 · similarity 0.86)
- Unraveling Batch Normalization for Realistic Test-Time Adaptation (AAAI 2024 · similarity 0.86)
- In-Place Test-Time Training (ICLR 2026 · similarity 0.85)
- ITP: Instance-Aware Test Pruning for Out-of-Distribution Detection (AAAI 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.
