Conventional fairness in multi-tenant Large Language Model (LLM) inference services is typically defined by system-centric metrics such as equitable resource allocation. We argue that this is unilateral and it creates a gap between measured system performance and actual user-perceived quality. We challenge this notion by introducing and formalizing Experiential Fairness, a user-centric paradigm that shifts the objective from equality of opportunity (resource access) to equity of outcome (user experience). With this motivation we propose ExFairS, a lightweight scheduling framework that perceives each user's satisfaction as a composite measure of Service Level Objective (SLO) compliance and resource consumption, and dynamically re-orders the serving queue guided by a credit-based priority mechanism. Extensive experiments on an 8-GPU NVIDIA V100 node show that ExFairS reduces the SLO violation rate by up to 100% and improves system throughput by 14-21.9%, outperforming state-of-the-art schedulers and delivering a demonstrably higher degree of Experiential Fairness.

To address the communication challenges associated with Federated Learning (FL), Decentralized Federated Learning (DFL) eliminates the central server and trains the model with decentralized method, enabling each client to only communicate with its neighbors. However, per our analysis, model trained with DFL experiences performance degradation because of data-heterogeneity and time-varying topologies. To address these issues, we propose a Dynamic K-step Gradient Tracking (DKGT) method to enhance the performance of DFL over time varying networks. Specifically, DKGT employs K-step local updates and gradient tracking to reduce the communication cost and the variance from heterogeneous data distribution, and we use dynamic gradient tracking parameter to correct gradient over time varying graph. Theoretically, we derive a universal convergence rate for smooth and non-convex problem at the rate of$\mathcal{O}\left(\frac{\left(f(\textbf{x}_0)-f(\textbf{x}^*)\right)}{\sqrt{T}(L\sqrt{KN})^{-1}-\tau(pKL\sqrt{TKN})^{-1}}+\frac{\sigma^2}{KTN\tau(pK-\tau)}\right)$, that τ and p respectively represent the time window length and the connectivity of time-varying networks. Experimentally, we illustrate the robustness and effectiveness of this heterogeneity correction on extensive non-convex neural network training tasks over different topologies and dynamic network settings.

Harmful fine-tuning attack poses a serious threat to the online fine-tuning service. Vaccine, a recent alignment-stage defense, applies uniform perturbation to all layers of embedding to make the model robust to the simulated embedding drift. However, applying layer-wise uniform perturbation may lead to excess perturbations for some particular non-safety-critical layers, resulting in defense performance degradation and unnecessary memory consumption. To address this limitation, we propose a Targeted Vaccine (T-Vaccine), a memory-efficient safety alignment method that applies perturbation to only selected layers of the model. T-Vaccine follows two core steps: First, it uses the harmful gradient norm as a statistical metric to identify the safety-critical layers. Second, instead of applying uniform perturbation across all layers, T-Vaccine only applies perturbation to the safety-critical layers while keeping other layers frozen during training. Results show that T-Vaccine outperforms Vaccine in terms of both defense effectiveness and resource efficiency. Comparison with other defense baselines, e.g., RepNoise and TAR also demonstrate the superiority of T-Vaccine. Notably, T-Vaccine is the first defense that enables a fine-tuning-based alignment method for 7B pre-trained models trained on consumer GPUs with limited memory (e.g., RTX 4090).

Federated learning (FL) is a distributed machine learning paradigm that enables heterogeneous devices to train a model collaboratively. Recognizing communication as a bottleneck in FL, existing communication-efficient solutions, e.g., HeteroFL and LotteryFL, etc., utilize gradient sparsification to reduce communication costs. However, existing solutions fail to address the dynamic bandwidth issue in which the bandwidth of each client is constantly changing throughout the training process. In this article, we propose AdaptiveFL, a communication-adaptive FL framework, considering the dynamic constraints of bandwidth. The design of AdaptiveFL follows two key steps: 1) in each round, each device selects a best-fit sub-model for communication per currently available bandwidth; and 2) to guarantee the performance of each sub-model sent under dynamic bandwidth constraints, AdaptiveFL employs a local training method that enables each device to train a "tailorable" local model, which can be tailored to any sparsity with competitive accuracy. We compare AdaptiveFL with several communication-efficient SOTA methods and demonstrate that AdaptiveFL outperforms other baselines by a large margin.

As deep learning (DL) workloads scale in complexity and volume, ensuring predictable job queuing times has become a critical challenge for data centers. Existing scheduling solutions primarily focus on minimizing tardiness or job completion times (JCT), often neglecting the need for deterministic queuing, particularly in dynamic and preemptive environments. This paper introducesKairos, a preemption-based scheduling framework enhanced by user collaboration to address these gaps.Kairoscombines adivide-and-conquerstrategy—segmenting jobs into sequential units with adaptive priorities—and a user-collaborative mechanism for better duration estimation. By leveraging real-time feedback from resource contention and queuing delays,Kairosminimizes a novel metric, theQueue inStability Index(QSI), achieving significant improvements in queuing predictability while maintaining competitive JCT. Experimental results demonstrate thatKairosreduces QSI by over 99.8% compared to state-of-the-art deadline-aware baselines, offering robust performance for diverse DL workloads.

Abstract Microservices architecture provides a promising solution for developing sustainable cyber‐physical systems (CPSs). However, checking the compatibility of CPSs over a set of microservices communicating asynchronously via unbounded buffers are undecidable due to their infinite state spaces. In this article, we propose a new approach for checking the compatibility of CPSs with infinite state spaces without restricting the size of buffers or the number of communication cycles. First, we integrate CPSs with microservice architecture and design the system architecture for building CPSs over a set of cyber‐physical microservices with unbounded buffers. Second, we model CPSs composed of asynchronously communicating cyber‐physical microservices via FIFO buffers as labelled transition systems. Third, we adopt the stability notion and present a sufficient condition for checking the unspecified receptions of CPSs through stability checking. Finally, we implement our approach in Process Analysis Toolkit for automatic compatibility checking and conduct experiments to show our approach is effective and efficient.