Research map / Label-efficient and robust learning / Test-time adaptation
TiME: Test-Time Mixture-of-Experts Routing via Asymmetric CO-Optimal Transport for Continual Test-Time Adaptation
Tianlun Liu, Zhiliang Tian, Zhen Huang, Tianle Liu, Xingzhi Zhou, Feng Liu, Dongsheng Li
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 TiME routing (OR variant) on top of best baseline CTTA-T | Qwen1.5-MoE-A2.7B · TU stream avg | average score: 0.4548 -> 0.4707 | Improved |
| add TiME (OR) over CTTA-T on DeepSeek | DeepSeek-V2-Lite · TU stream avg | average score: 0.4300 -> 0.4481 | Improved |
| use As-COOT instead of COOT (ablation) | TiME · TU stream avg (Qwen) | average score: 0.4553 -> 0.4707 | Improved |
| use As-COOT instead of Euclidean-distance allocation (ablation) | TiME · TU stream avg (Qwen) | average score: 0.4386 -> 0.4707 | Improved |
| use As-COOT instead of Hash Layer routing (ablation) | TiME · TU stream avg (Qwen) | average score: 0.4328 -> 0.4707 | Improved |
| add test-time online routing over original MoE routing (ablation) | TiME · TU stream avg (Qwen) | average score: 0.4354 -> 0.4707 | Improved |
| add semantic space alignment (SSA) (ablation) | TiME · TU stream avg (Qwen) | average score: 0.4517 -> 0.4707 | Improved |
| use Euclidean instead of cosine distance in SSA (ablation) | TiME · TU stream avg (Qwen) | average score: 0.4687 -> 0.4707 | No clear effect |
| use random restoration instead of L2 regularization (ablation) | TiME · TU stream avg (Qwen) | average score: 0.4641 -> 0.4707 | Improved |
| add random restoration of parameters (ablation) | TiME · TU stream avg (Qwen) | average score: 0.4534 -> 0.4707 | Improved |
| EOE vs OR compatibility construction in SSA (ablation) | TiME · TU stream avg (Qwen) | average score: 0.4661 vs 0.4707 | No clear effect |
| MoE base model instead of dense under Tent | Tent · TU stream avg | average score: 0.3543 -> 0.3887 | 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.
- SoTTA — SoTTA: Robust Test-Time Adaptation on Noisy Data Streams (NeurIPS 2023) · better on 2 of 2 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
- BECoTTA: Input-dependent Online Blending of Experts for Continual Test-time Adaptation (ICML 2024 · similarity 0.92)
- Class-aware Domain Knowledge Fusion and Fission for Continual Test-Time Adaptation (NeurIPS 2025 · similarity 0.88)
- ReservoirTTA: Prolonged Test-time Adaptation for Evolving and Recurring Domains (NeurIPS 2025 · similarity 0.87)
- DPCore: Dynamic Prompt Coreset for Continual Test-Time Adaptation (ICML 2025 · similarity 0.87)
- Towards Real-World Test-Time Adaptation: Tri-net Self-Training with Balanced Normalization (AAAI 2024 · similarity 0.86)
- A Versatile Framework for Continual Test-Time Domain Adaptation: Balancing Discriminability and Generalizability (CVPR 2024 · similarity 0.86)
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.
