The field of test-time adaptation is moving towards addressing the challenges of distribution shifts and domain adaptation in real-world scenarios. Researchers are exploring innovative methods to enhance the generalization of deep learning models, such as feature redundancy elimination, cross-domain diffusion, and reservoir-based adaptation. These approaches aim to improve the adaptability and robustness of models in the presence of evolving and recurring domains. Notable papers in this area include ReservoirTTA, which introduces a novel plug-in framework for prolonged test-time adaptation, and GCAL, which proposes a bilevel optimization strategy for graph domain adaptation. Additionally, ranked entropy minimization and cross-domain diffusion with progressive alignment are being investigated as effective methods for continual test-time adaptation and efficient adaptive retrieval.