Vision Foundation Models, such as DINO [20], have demonstrated remarkable generalization in classification; however, their application to out-of-domain visual regression tasks remains a significant and underexplored challenge. Unlike classification, domain generalization in regression poses distinct challenges: regression produces continuous outputs and is particularly sensitive to high-variance, label-irrelevant factors (e.g., illumination, blur, or contrast). These factors can entangle with task-relevant features and induce spurious correlations. While recent regression methods [11], [15], [24], [38], [39] have shown promise, they often rely on CNN backbones and require the pre-specification of known distractors. This demands significant domain expertise and fails to address spurious correlations that emerge during training. To address these challenges, we propose LDSC, a Latent Distractor Subspace Consistency framework that disentangles intermediate feature representation into task-relevant and latent distractor subspaces, and regularizes the latter under photometric perturbations to suppress spurious correlations while preserving discriminative features during training. Our proposed method, LDSC, is the first to effectively adapt the powerful DINO backbone for domain generalized visual regression. LDSC achieves state-of-the-art results on seven benchmark regression datasets, demonstrating its strong performance in domain generalization for visual regression with percentage improvements of (41.75%, 20.12%, 52.05%, 8.27%, 22.21%, 3.55%) over state-of-the-art DG regression methods, respectively. Project page is available: ldsc-iitd.github.io.

Function-as-a-service (FaaS) is a popular serverless computing paradigm for event-driven functions that elastically scale on public clouds. FaaS workflows (e.g.,AWS Step FunctionsandAzure Durable Functions), are composed from FaaS functions (e.g., AWS Lambda and Azure Functions) to build practical applications. But, the complex interactions between functions in the workflow and limited visibility into the internals of proprietary FaaS platforms are major impediments to analyzing a FaaS workflow's performance. While several works characterize FaaS platforms to derive such insights, or offer FaaS Workflow benchmarks, there is a lack of a principled of FaaS workflow platforms, which have unique scaling, performance and costing behavior influenced by the platform design, dataflow and workloads. In this article, we perform extensive evaluations of three popular FaaS workflow platforms from AWS and Azure, running 25 micro-benchmark and application workflows over$139k$invocations. Our detailed analysis confirms some conventional wisdom but also uncovers unique insights on the function execution, workflow orchestration, inter-function interactions, cold-start scaling and monetary costs. Our observations help developers better configure and program these platforms, set performance and scalability expectations, and identify research gaps on enhancing the platforms.

Functions-as-a-Service (FaaS) is a widely used serverless computing abstraction that helps developers build applications using event-driven, stateless functions that execute on the cloud. Commercial FaaS platforms such as AWS Lambda and Azure Functions offer elastic auto-scaling and invocation-level billing to ease operations. Applications are often composed as a dataflow of FaaS functions that are orchestrated by FaaS workflow platforms such as AWS Step Functions or Azure Durable Functions. However, the proprietary nature of FaaS platforms on public clouds means that their internals are less understood. While benchmarks to characterize FaaS platforms exist, none are available for a principled evaluation of FaaS workflow platforms. Further, they are less configurable and often limited to simple workloads and a single cloud provider. We address this limitation by proposing XFBench, an end-to-end automated benchmarking framework for FaaS workflows that works across clouds, and an accompanying function, workflow, and workload suite. The user provides a generic definition of the workflow and workload for benchmarking, and XFBench automatically deploys the workflows across multiple cloud platforms, generates client requests, and profiles the execution. We validate XFBench with realistic workflows and workloads on AWS and Azure platforms in different global regions to offer early insights into understanding the inter-function communication, function execution time, and cold start scaling.

Generalization on unseen domains is critical for Deep Neural Networks (DNNs) to perform well in real-world applications such as autonomous navigation. However, catastrophic forgetting limits the ability of domain generalization and unsupervised domain adaption approaches to adapt to constantly changing target domains. To overcome these challenges, We propose DoSe framework, a Domain-aware Self-Distillation method based on batch normalization prototypes to facilitate continual model generalization across varying target domains. Specifically, we enforce the consistency of batch normalization statistics between two batches of images sampled from the same target domain distribution between the student and teacher models. To alleviate catastrophic forgetting, we introduce a novel exemplar-based replay buffer to identify difficult samples for the model to retain the knowledge. Specifically, we demonstrate that identifying difficult samples and updating the model periodically using them can help in preserving knowledge learned from previously seen domains. We conduct extensive experiments on two real-world datasets ACDC, C-Driving, and one synthetic dataset SHIFT to verify the efficiency of the proposed DoSe framework. On ACDC, our method outperforms existing SOTA in Domain Generalization, Unsupervised Domain Adaptation, and Daytime settings by 26%, 14%, and 70% respectively.