Test-time prompt tuning (TPT) aims to adjust the vision-language models (e.g., CLIP) with learnable prompts during the inference phase. However, previous works overlooked that pre-trained models as a service (MaaS) have become a noticeable trend due to their commercial usage and potential risk of misuse. In the context of MaaS, users can only design prompts in inputs and query the black-box vision-language models through inference APIs, rendering the previous paradigm of utilizing gradient for prompt tuning is infeasible. In this paper, we propose black-box test-time prompt tuning (B²TPT), a novel framework that addresses the challenge of optimizing prompts without gradients in an unsupervised manner. Specifically, B²TPT designs a consistent or confident (CoC) pseudo-labeling strategy to generate high-quality pseudo-labels from the outputs. Subsequently, we propose to optimize low-dimensional intrinsic prompts using a derivative-free evolution algorithm and to project them onto the original text and vision prompts. This strategy addresses the gradient-free challenge while reducing complexity. Extensive experiments across 15 datasets demonstrate the superiority of B²TPT. The results show that B²TPT not only outperforms CLIP's zero-shot inference at test time, but also surpasses other gradient-based TPT methods.

Efficient container scaling is crucial for enhancing the availability and scalability of cloud-native applications through adaptive resource management. In cloud computing, the default autoscaling feature of Kubernetes scales pods only when the cluster or application exceeds a predefined threshold. However, this reactive approach often leads to significant resource waste during demand fluctuations because it cannot predict future workload changes and adjust resources in advance. This paper presents CATScaler, a novel Convolution-Augmented Transformer Scaler designed to proactively optimize resource allocation in serverless environments. CATScaler is a proactive approach composed of two modules: workload prediction and elastic auto-scaling. In the prediction module, we develop a convolution-augmented transformer to accurately predict workload changes at both local and global levels. Additionally, we incorporate reversible instance normalization to mitigate the shift caused by the difference between workload data and training data. In the auto-scaling module, we implement an instance-counting method to handle the nonlinear relationships between variables. Experiments using two real datasets from Alibaba Cloud and Huawei Cloud demonstrate the effectiveness of CATScaler. The tests conducted on a cluster of 4 servers demonstrated that CATScaler reduced response time latency by 1.1× compared to Kubernetes' default scaler and decreased service violation rates by 3.2×.

Unsupervised domain adaptation aims to transfer the knowledge learned from a labeled source domain to an unlabeled target domain with different data distributions. However, in practice, source samples are not always available due to privacy protection and storage resource limitations. To address this concern, Source-Free Domain Adaptation (SFDA) has recently attracted growing research attention, as it only needs a pre-trained source model without direct access to source data. In this paper, we propose a novel Adversarial SOurce GEneration (ASOGE) method for SFDA, which introduces an additional generative module to produce synthetic labeled source samples and uses them to facilitate cross-domain adaptation. Unlike early studies that train the generator independently and perform the adaptation only after the generator is finished, ASOGE integrates the generation and adaptation stages within a collaborative framework by making them play an adversarial game. In the generation stage, the labeled source samples are not produced blindly; instead, they are hard-to-align samples that provide knowledge more worth learning for the adaptation stage. To achieve a fine-grained domain alignment, a class-aware discrepancy between source and target domains is measured via contrastive learning. Extensive experiments on benchmark datasets demonstrate the effectiveness of ASOGE compared to the state-of-the-art methods.