In the rapidly evolving landscape of 6G wireless networks, time-sensitive applications demand efficient data transmission that prioritizes meaningful content over raw signals. This paper introduces a novel adaptive semantic communication system designed for ultra-reliable and low-latency applications, such as autonomous vehicles, drone monitoring, and satellite imaging. The system uses a multi-exit autoencoder to dynamically alter encoding depth based on input data complexity, maximizing the balance of semantic accuracy and processing performance. A lightweight contextual multi-armed bandit (CMAB) mechanism guides exit selection, leveraging a hybrid feature extractor that combines an encoder-derived semantic features vector with statistical image metrics to enable real-time decisions without supervision. The design of this adaptable architecture, an online learning policy for resource allocation, and validation through tests with object detection benchmarks are all significant contributions. Results demonstrate superior performance over traditional and neural-based codecs, particularly under high compression ratios, with enhanced fidelity in latency-constrained environments. This approach paves the way for practical 6G deployments in edge-intelligent systems, offering reduced overhead while preserving critical information for upstream and downstream tasks.

Distributed quantum computing (DQC) mitigates the hardware limitations of near-term devices by partitioning large circuits into smaller segments executed on qubit-limited processors. However, even these reduced circuits often exceed the capacity of near-term devices that provide only extremely few reliable qubits. To address this, we introduce a qubit-reuse based DQC framework that exploits mid-circuit measurement and reset to enable scalable execution with minimal physical qubits. We consider distributed quantum learning (DQL) as a representative case study and formulate an energy minimization problem for the proposed framework that jointly optimizes quantum operation energy, measurement repetitions, and transmission power. To tackle the inherent non-convexity, we design a hybrid block coordinate descent (BCD) and successive convex approximation (SCA) algorithm for efficient solution. Extensive simulations on MNIST and CIFAR-10 demonstrate that the proposed qubit-reuse framework enables scalable execution on qubit-limited devices while achieving competitive performance compared to centralized quantum computing, and that the joint optimization algorithm significantly improves energy efficiency over baseline schemes.

Distributed quantum learning (DQL) has emerged as a promising paradigm to scale quantum-enhanced machine learning by interconnecting multiple quantum devices. However, for efficient real-world deployment, it is essential to characterize how DQL converges under practical scenarios while simultaneously safeguarding multi-device quantum infrastructures from evolving security threats. Addressing these aspects in an integrated manner is key to ensuring both performance and resilience in large-scale DQL systems. Therefore, this paper presents a new DQL study where our innovation lies in: (i) conducting a holistic convergence analysis for DQL under practical settings, i.e., partial device participation, non-convex loss functions, and heterogeneous data distributions, (ii) developing a novel multi-layered post-quantum cryptographic architecture with a quantum neural network-powered adaptive mechanism that monitors conditions, evaluates threats, and adjusts parameters across three National Institute of Standards and Technology (NIST)-compliant levels. Our theoretical framework and empirical validation reveal two key insights: (i) the derived convergence bound uncovers a fundamental trade-off between convergence rate, measurement shots, and the size of the participating device subset; and (ii) findings from our evaluations on a physical testbed modeling quantum control architectures expose the performance limitations of static post-quantum security, while confirming that our adaptive framework effectively mitigates these overheads to preserve overall system efficiency. Specifically, the hardware experiments demonstrate that our dynamic security mechanism reduces total security execution time by approximately 49% relative to static high-security baselines, while maintaining a threat detection accuracy of over 91%. Furthermore, extensive simulations validate our theoretical analysis, showing strong agreement between predicted and observed convergence trends. The coupling of these two stages ensures both theoretically grounded performance optimization and adaptive threat resilience, enabling efficient, secure, and scalable DQL deployment.