Searcharxiv⌕ Search

arXiv subjects

Kun Yang

Publications and source records attributed to Kun Yang.

At least 55 records · Page 3Linked to original sources
arXiv2026

Agentic AI-Empowered Wireless Agent Networks With Semantic-Aware Collaboration via ILAC

The rapid development of agentic artificial intelligence (AI) is driving future wireless networks to evolve from passive data pipes into intelligent collaborative ecosystems under the emerging paradigm of integrated learning and communication (ILAC). However, realizing efficient agentic collaboration faces challenges not only in handling semantic redundancy but also in the lack of an integrated mechanism for communication, computation, and control. To address this, we propose a wireless agent network (WAN) framework that orchestrates a progressive knowledge aggregation mechanism. Specifically, we formulate the aggregation process as a joint energy minimization problem where the agents perform semantic compression to eliminate redundancy, optimize transmission power to deliver semantic payloads, and adjust physical trajectories to proactively enhance channel qualities. To solve this problem, we develop a hierarchical algorithm that integrates inner-level resource optimization with outer-level topology evolution. Theoretically, we reveal that incorporating a potential field into the topology evolution effectively overcomes the short-sightedness of greedy matching, providing a mathematically rigorous heuristic for long-term energy minimization. Simulation results demonstrate that the proposed framework achieves superior energy efficiency and scalability compared to conventional benchmarks, validating the efficacy of semantic-aware collaboration in dynamic environments.

cs.NI↗
arXiv2026

CoDec: Prefix-Shared Decoding Kernel for LLMs

Prefix-sharing among multiple prompts presents opportunities to combine the operations of the shared prefix, while attention computation in the decode stage, which becomes a critical bottleneck with increasing context lengths, is a memory-intensive process requiring heavy memory access on the key-value (KV) cache of the prefixes. Therefore, in this paper, we explore the potential of prefix-sharing in the attention computation of the decode stage. However, the tree structure of the prefix-sharing mechanism presents significant challenges for attention computation in efficiently processing shared KV cache access patterns while managing complex dependencies and balancing irregular workloads. To address the above challenges, we propose a dedicated attention kernel to combine the memory access of shared prefixes in the decoding stage, namely CoDec. CoDec delivers two key innovations: a novel shared-prefix attention kernel that optimizes memory hierarchy and exploits both intra-block and inter-block parallelism, and a comprehensive workload balancing mechanism that efficiently estimates cost, divides tasks, and schedules execution. Experimental results show that CoDec achieves an average $1.9\times$ speedup and $120.9\times$ memory access reduction compared to the state-of-the-art FlashDecoding kernel regarding attention computation in the decode stage and $3.8\times$ end-to-end time per output token compared to the vLLM.

cs.LG↗
arXiv2026

Effect of edge-stretching on Steklov eigenvalues and sharp Steklov eigenvalue bounds on leaf--boundary trees

Let $T$ be a finite tree with leaf set $\dO$ as the boundary and let $λ_2$ be the first nontrivial Steklov eigenvalue. Let $D$ and $\ell$ be the maximum vertex degree and the number of leaves, respectively. Motivated by the spectral influence of neck-stretching on Riemannian manifolds, we investigate a discrete counterpart--edge-stretching--and its effect on the Steklov eigenvalues of graphs. We prove that Steklov eigenvalues decrease monotonically under the edge--stretching operation. As a consequence, we prove that $λ_2\le D/\ell$, with equality if and only if $T$ is a star. This fundamentally improves the constant in He--Hua's bound $λ_2\le 4(D-1)/\ell$ to the optimal value~$1$. We also provide a closed-form diagonalization of the Steklov problem on level--regular trees, yielding explicit eigenvalues and multiplicities. In addition, we provide a general upper bound $λ_k\le \min\{1,\,16Dk/\ell\}$ for higher eigenvalues. Systematic numerical experiments verify the sharp bound and provide evidence for the extremal conjecture of Lin--Zhao on balanced minimum--height trees.

math.CO↗
arXiv2026

LoD-Loc v3: Generalized Aerial Localization in Dense Cities using Instance Silhouette Alignment

We present LoD-Loc v3, a novel method for generalized aerial visual localization in dense urban environments. While prior work LoD-Loc v2 achieves localization through semantic building silhouette alignment with low-detail city models, it suffers from two key limitations: poor cross-scene generalization and frequent failure in dense building scenes. Our method addresses these challenges through two key innovations. First, we develop a new synthetic data generation pipeline that produces InsLoD-Loc - the largest instance segmentation dataset for aerial imagery to date, comprising 100k images with precise instance building annotations. This enables trained models to exhibit remarkable zero-shot generalization capability. Second, we reformulate the localization paradigm by shifting from semantic to instance silhouette alignment, which significantly reduces pose estimation ambiguity in dense scenes. Extensive experiments demonstrate that LoD-Loc v3 outperforms existing state-of-the-art (SOTA) baselines, achieving superior performance in both cross-scene and dense urban scenarios with a large margin. The project is available at https://nudt-sawlab.github.io/LoD-Locv3/.

cs.CV↗
arXiv2026

A Unified Cloud-Edge-Terminal Framework for Multimodal Integrated Sensing and Communication

The transition to 6G calls for tightly integrated sensing and communication to support mission-critical services such as autonomous driving, embodied AI, and high-precision telemedicine. However, most existing ISAC designs rely on a single sensing modality (often RF), which limits environmental understanding and becomes a bottleneck in complex and dynamic scenes. This motivates a shift from single-modal to multimodal ISAC, where heterogeneous sensors (e.g., radar, LiDAR, and cameras) complement each other to improve robustness and semantic awareness. In this article, we first summarize key challenges for multimodal ISAC, including heterogeneous fusion, communication overhead, and scalable system design. We then highlight three enabling technologies: large AI models, semantic communications, and multi-agent systems, and discuss how their combination can enable task-oriented multimodal perception. Building on these insights, we propose a unified cloud-edge-terminal (CET) framework that hierarchically distributes intelligence and supports three adaptive operation modes: global fusion mode (GFM), cooperative relay mode (CRM), and peer interaction mode (PIM). A case study evaluates the framework across three modes, demonstrating that GFM achieves the highest accuracy, PIM minimizes latency, and CRM strikes an optimal balance between performance and efficiency. Finally, we conclude with open research issues and future directions.

cs.NI↗
arXiv2026

Immersive Volumetric Video Playback: Near-RT Resource Allocation and O-RAN-based Implementation

Immersive volumetric video streaming in extended reality (XR) demands ultra-low motion-to-photon (MTP) latency, which conventional edge-centric architectures struggle to meet due to per-frame computationally intensive rendering tightly coupled with user motion. To address this challenge, we propose an Open Radio Access Network (O-RAN)-integrated playback framework that jointly orchestrates radio, compute, and content resources in near real time (Near-RT) control loop. The system formulates the rendered-pixel ratio as a continuous control variable and jointly optimizes it over the Open Cloud (O-Cloud) compute, gNB transmit power, and bandwidth under a Weber-Fechner quality of experience (QoE) model, explicitly balancing resolution, computation, and latency. A Soft Actor-Critic (SAC) agent with structured action decomposition and QoE-aware reward shaping resolves the resulting high-dimensional control problem. Experiments on a 5G O-RAN testbed and system simulations show that SAC reduces median MTP latency by above $11\%$ and improves both mean QoE and fairness, demonstrating the feasibility of RIC-driven joint radio-compute-content control for scalable, latency-aware immersive streaming.

cs.NI↗
arXiv2026

Edge-Cloud Collaborative Computing on Distributed Intelligence and Model Optimization: A Survey

Edge-cloud collaborative computing (ECCC) has emerged as a pivotal paradigm for addressing the computational demands of modern intelligent applications, integrating cloud resources with edge devices to enable efficient, low-latency processing. Recent advancements in AI, particularly deep learning and large language models (LLMs), have dramatically enhanced the capabilities of these distributed systems, yet introduce significant challenges in model deployment and resource management. In this survey, we comprehensive examine the intersection of distributed intelligence and model optimization within edge-cloud environments, providing a structured tutorial on fundamental architectures, enabling technologies, and emerging applications. Additionally, we systematically analyze model optimization approaches, including compression, adaptation, and neural architecture search, alongside AI-driven resource management strategies that balance performance, energy efficiency, and latency requirements. We further explore critical aspects of privacy protection and security enhancement within ECCC systems and examines practical deployments through diverse applications, spanning autonomous driving, healthcare, and industrial automation. Performance analysis and benchmarking techniques are also thoroughly explored to establish evaluation standards for these complex systems. Furthermore, the review identifies critical research directions including LLMs deployment, 6G integration, neuromorphic computing, and quantum computing, offering a roadmap for addressing persistent challenges in heterogeneity management, real-time processing, and scalability. By bridging theoretical advancements and practical deployments, this survey offers researchers and practitioners a holistic perspective on leveraging AI to optimize distributed computing environments, fostering innovation in next-generation intelligent systems.

cs.DC↗
arXiv2026

OOD-MMSafe: Advancing MLLM Safety from Harmful Intent to Hidden Consequences

While safety alignment for Multimodal Large Language Models (MLLMs) has gained significant attention, current paradigms primarily target malicious intent or situational violations. We propose shifting the safety frontier toward consequence-driven safety, a paradigm essential for the robust deployment of autonomous and embodied agents. To formalize this shift, we introduce OOD-MMSafe, a benchmark comprising 455 curated query-image pairs designed to evaluate a model's ability to identify latent hazards within context-dependent causal chains. Our analysis reveals a pervasive causal blindness among frontier models, with the highest 67.5% failure rate in high-capacity closed-source models, and identifies a preference ceiling where static alignment yields format-centric failures rather than improved safety reasoning as model capacity grows. To address these bottlenecks, we develop the Consequence-Aware Safety Policy Optimization (CASPO) framework, which integrates the model's intrinsic reasoning as a dynamic reference for token-level self-distillation rewards. Experimental results demonstrate that CASPO significantly enhances consequence projection, reducing the failure ratio of risk identification to 7.3% for Qwen2.5-VL-7B and 5.7% for Qwen3-VL-4B while maintaining overall effectiveness.

cs.AI↗
arXiv2026

Spherical Latent Motion Prior for Physics-Based Simulated Humanoid Control

Learning motion priors for physics-based humanoid control is an active research topic. Existing approaches mainly include variational autoencoders (VAE) and adversarial motion priors (AMP). VAE introduces information loss, and random latent sampling may sometimes produce invalid behaviors. AMP suffers from mode collapse and struggles to capture diverse motion skills. We present the Spherical Latent Motion Prior (SLMP), a two-stage method for learning motion priors. In the first stage, we train a high-quality motion tracking controller. In the second stage, we distill the tracking controller into a spherical latent space. A combination of distillation, a discriminator, and a discriminator-guided local semantic consistency constraint shapes a structured latent action space, allowing stable random sampling without information loss. To evaluate SLMP, we collect a two-hour human combat motion capture dataset and show that SLMP preserves fine motion detail without information loss, and random sampling yields semantically valid and stable behaviors. When applied to a two-agent physics-based combat task, SLMP produces human-like and physically plausible combat behaviors only using simple rule-based rewards. Furthermore, SLMP generalizes across different humanoid robot morphologies, demonstrating its transferability beyond a single simulated avatar.

cs.RO↗
arXiv2026

Pragma-VL: Towards a Pragmatic Arbitration of Safety and Helpfulness in MLLMs

Multimodal Large Language Models (MLLMs) pose critical safety challenges, as they are susceptible not only to adversarial attacks such as jailbreaking but also to inadvertently generating harmful content for benign users. While internal safety alignment via Supervised Fine-Tuning (SFT) and Reinforcement Learning (RL) is a primary mitigation strategy, current methods often face a safety-utility trade-off: they either refuse benign queries out of excessive caution or overlook latent risks in cross-modal interactions. To resolve this, we introduce Pragma-VL, an end-to-end alignment algorithm that enables MLLMs to pragmatically arbitrate between safety and helpfulness. First, we enhance visual risk perception with a novel cold-start SFT stage. This is achieved by applying risk-aware clustering to the visual encoder and using an interleaved dataset of risk descriptions and high-quality data. Second, we introduce a theoretically-guaranteed reward model that leverages synergistic learning. We train it with a novel data augmentation method that assigns dynamic weights based on the queries, enabling contextual arbitration between safety and helpfulness. Extensive experiments show that Pragma-VL effectively balances safety and helpfulness, outperforming baselines by 5% to 20% on most multimodal safety benchmarks while preserving its general capabilities in areas such as mathematics and knowledge reasoning.

cs.LG↗
arXiv2026

ManCAR: Manifold-Constrained Latent Reasoning with Adaptive Test-Time Computation for Sequential Recommendation

Sequential recommendation increasingly employs latent multi-step reasoning to enhance test-time computation. Despite empirical gains, existing approaches largely drive intermediate reasoning states via target-dominant objectives without imposing explicit feasibility constraints. This results in latent drift, where reasoning trajectories deviate into implausible regions. We argue that effective recommendation reasoning should instead be viewed as navigation on a collaborative manifold rather than free-form latent refinement. To this end, we propose ManCAR (Manifold-Constrained Adaptive Reasoning), a principled framework that grounds reasoning within the topology of a global interaction graph. ManCAR constructs a local intent prior from the collaborative neighborhood of a user's recent actions, represented as a distribution over the item simplex. During training, the model progressively aligns its latent predictive distribution with this prior, forcing the reasoning trajectory to remain within the valid manifold. At test time, reasoning proceeds adaptively until the predictive distribution stabilizes, avoiding over-refinement. We provide a variational interpretation of ManCAR to theoretically validate its drift-prevention and adaptive test-time stopping mechanisms. Experiments on seven benchmarks demonstrate that ManCAR consistently outperforms state-of-the-art baselines, achieving up to a 46.88% relative improvement w.r.t. NDCG@10. Our code is available at https://github.com/FuCongResearchSquad/ManCAR.

cs.IR↗
arXiv2026

Resource Allocation for STAR-RIS-enhanced Metaverse Systems with Augmented Reality

Augmented reality (AR)-enabled Metaverse is a promising technique to provide immersive service experience for mobile users. However, the limited network resources and unpredictable wireless propagation environments are key design bottlenecks of AR-enabled Metaverse systems. Therefore, this paper presents a resource management framework for simultaneously transmitting and reflecting RIS (STAR-RIS)-assisted AR-enabled Metaverse, where the STAR-RIS is configured to improve the communication efficiency between AR users and the Metaverse server located at the base station (BS). Moreover, we formulate a service latency minimization problem via jointly optimizing the computation resource allocation of the BS, coefficient matrix of the STAR-RIS, central processing unit (CPU) frequency and transmit power of the AR users. To tackle the non-convex problem, we utilize an approximate method to transform it to a tractable form, and decouple the multi-dimensional variables via the alternating optimization method. Particularly, the optimal coefficient matrix is obtained by a penalty function-based method with proved convergence, the CPU frequencies of AR users are derived as the closed-form solution, and the transmit power of AR users and computation resource allocation of the BS are obtained by the Lagrange duality method and convex optimization theory. Finally, simulation results demonstrates that the proposed method achieves remarkable latency reduction than several benchmark methods.

cs.IT↗
arXiv2026

Omni-Safety under Cross-Modality Conflict: Vulnerabilities, Dynamics Mechanisms and Efficient Alignment

Omni-modal Large Language Models (OLLMs) greatly expand LLMs' multimodal capabilities but also introduce cross-modal safety risks. However, a systematic understanding of vulnerabilities in omni-modal interactions remains lacking. To bridge this gap, we establish a modality-semantics decoupling principle and construct the AdvBench-Omni dataset, which reveals a significant vulnerability in OLLMs. Mechanistic analysis uncovers a Mid-layer Dissolution phenomenon driven by refusal vector magnitude shrinkage, alongside the existence of a modal-invariant pure refusal direction. Inspired by these insights, we extract a golden refusal vector using Singular Value Decomposition and propose OmniSteer, which utilizes lightweight adapters to modulate intervention intensity adaptively. Extensive experiments show that our method not only increases the Refusal Success Rate against harmful inputs from 69.9% to 91.2%, but also effectively preserves the general capabilities across all modalities. Our code is available at: https://github.com/zhrli324/omni-safety-research.

cs.CR↗
arXiv2026

A Novel Hierarchical Co-Optimization Framework for Coordinated Task Scheduling and Power Dispatch in Computing Power Networks

The proliferation of large-scale AI and data-intensive applications has driven the development of Computing Power Networks (CPN). It is a key paradigm for delivering ubiquitous, on-demand computational services with high efficiency. However, CPNs face dual challenges in service computing. Immense energy consumption threatens sustainable operations. And the integration with power grids also features high penetration of intermittent Renewable Energy Sources (RES), complicating task scheduling while ensuring Quality of Service (QoS). To address these issues, this paper proposes a novel Two-Stage Co-Optimization (TSCO) framework. It synergistically coordinates CPN task scheduling and power system dispatch, aiming to optimize service performance while achieving low-carbon operations. The framework decomposes the complex, large-scale problem into a day-ahead stochastic unit commitment stage and a real-time operational stage. The former is solved using Benders decomposition for computational tractability, while in the latter, economic dispatch of generation assets is coupled with an adaptive CPN task scheduling managed by a deep reinforcement learning agent. It makes carbon-aware decisions by responding to dynamic grid conditions, including real-time electricity prices and marginal carbon intensity. Extensive simulations demonstrate that the TSCO outperforms baseline approaches significantly. It reduces carbon emissions by 16.2% and operational costs by 12.7%, while decreasing RES curtailment by over $60\%$, maintaining a task success rate of 98.5%, and minimizing average task tardiness to 12.3s. This work advances cross-domain service optimization in CPNs.

cs.NI↗
arXiv2026

CSR-Bench: A Benchmark for Evaluating the Cross-modal Safety and Reliability of MLLMs

Multimodal large language models (MLLMs) enable interaction over both text and images, but their safety behavior can be driven by unimodal shortcuts instead of true joint intent understanding. We introduce CSR-Bench, a benchmark for evaluating cross-modal reliability through four stress-testing interaction patterns spanning Safety, Over-rejection, Bias, and Hallucination, covering 61 fine-grained types. Each instance is constructed to require integrated image-text interpretation, and we additionally provide paired text-only controls to diagnose modality-induced behavior shifts. We evaluate 16 state-of-the-art MLLMs and observe systematic cross-modal alignment gaps. Models show weak safety awareness, strong language dominance under interference, and consistent performance degradation from text-only controls to multimodal inputs. We also observe a clear trade-off between reducing over-rejection and maintaining safe, non-discriminatory behavior, suggesting that some apparent safety gains may come from refusal-oriented heuristics rather than robust intent understanding. WARNING: This paper contains unsafe contents.

cs.AI↗
arXiv2026

Superimposed-Pilot OTFS Under Fractional Doppler: Modular End-to-End Learning

Orthogonal time frequency space (OTFS) modulation has emerged as a promising candidate to overcome the performance degradation of orthogonal frequency division multiplexing (OFDM), which are commonly encountered in high-mobility wireless communication scenarios. However, conventional OTFS transceivers rely on multiple separately designed signal-processing modules, whose isolated optimization often limits global optimal performance. To overcome limitations, this paper proposes a modular deep learning (DL) based end-to-end OTFS transceiver framework that consists of trainable and interchangeable neural network (NN) modules, including constellation mapping/demapping, superimposed pilot placement, inverse Zak (IZak)/Zak transforms, and a U-Net-enhanced NN tailored for joint channel estimation and detection (JCED), while explicitly accounting for the impact of the cyclic prefix. This physics-informed modular architecture provides flexibility for integration with conventional OTFS systems and adaptability to different communication configurations. Simulations demonstrate that the proposed design significantly outperforms baseline methods in terms of both normalized mean squared error (NMSE) and detection reliability, maintaining robustness under integer and fractional Doppler conditions. The results highlight the potential of DL-based end-to-end optimization to enable practical and high-performance OTFS transceivers for next-generation high-mobility networks.

eess.SP↗
arXiv2026

Adaptive Requesting in Decentralized Edge Networks via Non-Stationary Bandits

We study a decentralized collaborative requesting problem that aims to optimize the information freshness of time-sensitive clients in edge networks consisting of multiple clients, access nodes (ANs), and servers. Clients request content through ANs acting as gateways, without observing AN states or the actions of other clients. We define the reward as the age of information reduction resulting from a client's selection of an AN, and formulate the problem as a non-stationary multi-armed bandit. In this decentralized and partially observable setting, the resulting reward process is history-dependent and coupled across clients, and exhibits both abrupt and gradual changes in expected rewards, rendering classical bandit-based approaches ineffective. To address these challenges, we propose the AGING BANDIT WITH ADAPTIVE RESET algorithm, which combines adaptive windowing with periodic monitoring to track evolving reward distributions. We establish theoretical performance guarantees showing that the proposed algorithm achieves near-optimal performance, and we validate the theoretical results through simulations.

cs.LG↗
arXiv2026

Extended Target Adaptive Beamforming for ISAC:A Perspective of Predictive Error Ellipse

Utilizing communication signals to extract motion parameters has emerged as a key direction in Vehicle-to- Everything (V2X) networks. Accurately modeling the relationship between communication signals and sensing performance is critical for the advancement of such systems. Unlike prior work that relies primarily on qualitative analysis, this paper derives the Cramér-Rao Bound (CRB) for radar parameter estimation in the context of Orthogonal Frequency Division Multiplexing (OFDM) waveforms and Uniform Planar Array (UPA) configurations. Recognizing that vehicles may act as extended targets, we propose two New Radio (NR)-V2X-compatible beamforming schemes tailored to different phases of the communication process. During the initial beam establishment phase, we develop a beamforming approach based on the union of predictive error ellipses, which enhances scatterer localization through temporally assisted beam training. In the beam adjustment phase, we introduce an adaptive narrowest-beam strategy that leverages the positions of scatterers and the communication receiver (CR), enabling effective tracking with reduced complexity. The beam design problem is addressed using the minimum enclosing ellipse algorithm and tailored antenna control methods. Simulation results validate the proposed approach, showing up to a 32.4% improvement in achievable rate with a 32*32 transmit antenna array and a 5.2% gain with an 8*8 array, compared to conventional beam sweeping under identical SNR conditions.

cs.IT↗