arXiv subjects

Haijun Zhang

Publications and source records attributed to Haijun Zhang.

At least 19 recordsLinked to original sources

arXiv2026

TrimMoE A communication aware and adaptive depth framework for distributed edge inference

Serving Mixture-of-Experts (MoE) large language models across distributed edge servers is bottlenecked by the cross-server expert transmission. The existing approaches mainly focus on how to reach a remote expert faster. However, in this paper, we instead consider whether a given layer, and the layers after it, need to be executed at all. To this end, a communication-aware adaptive-depth framework is proposed in this paper, termed TrimMoE, which couples layer skipping and confidence-based early exit with substitute execution and server-expert selection under a unified quality budget. Specifically, in the offline stage, TrimMoE freezes the backbone, trains the lightweight per-layer exit heads, calibrates the per-layer importance thresholds, and allocates the expert replicas by a skip/exit-aware redundancy benefit. In the online stage, a transition-aware look-ahead anticipates the token movement, so that the depth reduction targets the costliest transmissions, and besides, two feedback rules adapt the delay-quality weights and the exit threshold. Moreover, we prove that the substitution-and-skipping proxy degradation never exceeds the configured budget, and that the early exit is admitted only under a calibrated confidence gate. On a heterogeneous 10-server testbed with Switch-Base-8E, Qwen-MoE-A2.7B, and Mixtral-8x7B, TrimMoE reduces the average latency by up to 62.8%, lowers the cross-server traffic and the remote-execution ratio, and sustains high throughput under load, while keeping the task-quality degradation within a 2% bound.

cs.NI↗

arXiv2026

HetRoute Heterogeneous and Cost-aware Collaborative Routing Framework for Distributed Edge MoE Inference

Mixture-of-Experts (MoE) models have become a dominant architecture for large-scale AI services, yet deploying them over geo-distributed heterogeneous edge servers remains challenging. When the Top-k activated experts of a token are spread across multiple servers, the optimal routing depends jointly on cross-server link bandwidth, heterogeneous GPU computing capability, GPU-CPU expert loading delay, instantaneous queueing backlog, and replica-level quantization quality loss. Existing distributed inference and MoE serving methods address these factors separately and do not provide a unified framework for online multi-server collaborative routing. In this paper, we propose HetRoute, a heterogeneous-cost-aware collaborative routing framework for distributed edge MoE inference. HetRoute introduces a unified per-assignment cost model that explicitly captures four cost components: cross-server transmission, GPU-CPU offloading, GPU computation with queueing, and quantization-induced quality penalty. Guided by this model, the offline stage determines expert server placement, GPU-CPU residency, and replica precision through a routing-cost-coupled deployment algorithm, while the online stage routes the Top-k activated expert set as a whole by minimizing the bottleneck layer cost via exact enumeration or beam search. Theoretical analysis establishes fallback feasibility, a bound on the number of participating servers, per-layer optimality for small candidate domains, and online computational complexity. Trace-driven evaluation on three MoE models over a heterogeneous 10-server edge testbed shows that HetRoute reduces average inference latency by up to 59.0% and P99 latency by up to 58.0%, cuts cross-server traffic by up to 72.1%, and achieves 2.13x throughput improvement compared with representative baselines, while keeping quality degradation within the configured budget.

cs.NI↗

arXiv2026

Beamforming and Phase Shift Design for STAR-RIS Assisted Secure Sensing and Communication in ISAC Systems

Integrated sensing and communication(ISAC), as a rapidly advancing technique, introduces a fresh approach for achieving secure communication and intelligent sensing for future wireless networks. An ISAC framework empowered by simultaneously transmitting and reflecting reconfigurable intelligent surfaces(STAR-RIS) is explored in this paper, where a base station equipped with multiple antennas establishes wireless links to users each with a single antenna during the detection of a point target. The point target, regarded as an eavesdropper, trying to intercept users' information. Cramer-Rao bound(CRB) serves as evaluation criterion to assess sensing accuracy of point eavesdropper, whereas the secrecy rate is employed to quantify the security level of the communication link. To optimize sensing-communication tradeoff, a joint optimization problem is constructed. To approach the formulated problem, a hybrid Block Coordinate Descent(BCD)-based algorithm is developed, which alternately updates the transmission beamforming and STAR-RIS phase shifts, using successive convex approximation(SCA) technique, penalty dual decomposition (PDD) framework and projected gradient method(PGM).

eess.SP↗

arXiv2026

Harnessing Large Language Models for Intelligent Resource Allocation in the Internet of Everything

The rapid development of the Internet of Everything (IoE) is accelerating the adoption of intelligent applications. However, the massive number of connected devices generates diverse and heterogeneous tasks, which pose increasing challenges for dynamic resource scheduling in IoE environments. Using their superior semantic understanding and reasoning capabilities, Large Artificial Intelligence Models (LAIMs) demonstrate significant potential to handle complex scheduling scenarios and improve resource utilization efficiency. This paper investigates a task-oriented LAIM-driven resource scheduling mechanism, which constructs a multidimensional scheduling decision model by integrating task semantics, network states, and constraint conditions. Furthermore, a task-oriented prompt generation method is designed to establish a deep association between task requirements and network state. In the proposed resource allocation scheme, an external evaluation and feedback module is incorporated to conduct real-time feasibility verification and performance evaluation of scheduling strategies, thus enhancing the robustness and adaptability of scheduling. Simulation results demonstrate that the proposed Large Language Model (LLM)-driven network architecture and resource allocation scheme achieve significant improvements in convergence speed, processing latency, and energy consumption, effectively enhancing IoE task responsiveness and resource utilization.

cs.NI↗

arXiv2026

Energy Minimization Oriented Resource Allocation for Integrated Sensing and Communication in Marine IoT Networks

Integrated sensing and communication (ISAC) has become a promising technical framework for Marine Internet of Things (MIoT) systems. Nevertheless, all devices rely on battery power, so energy efficiency becomes a core bottleneck limiting practical deployment. This paper investigates the energy consumption minimization problem of MIoT-oriented ISAC systems. In this system, an uncrewed aerial vehicle (UAV) uses non-orthogonal multiple access (NOMA) to simultaneously perform target sensing and collect data from uncrewed surface vehicles (USVs), then forwards processed sensing information and USV data to a shore-based base station (SBS). Subject to latency limits and sensing performance requirements, total system energy consumption can be minimized via joint optimization of multiple variables, UAV transmit beamforming, dedicated sensing signal, USV transmit power, UAV computation power, and time resource allocation for sensing and communication phases. To tackle this non-convex optimization problem, we build a layered solution architecture that divides the original problem into independent subproblems and optimizes each alternately according to its mathematical features. Specifically, we first derive closed-form USV transmit power solutions and conduct variable substitution. The successive convex approximation (SCA) method is adopted to convert remaining non-convex subproblems into convex forms, on which we design efficient iterative algorithms. Simulation results verify the validity and accuracy of our algorithm in reducing system energy consumption. Compared with orthogonal frequency division multiple access (OFDMA) and genetic algorithm benchmarks, our scheme lowers system energy consumption by 19.71% and 8%, respectively. In addition, our optimized energy value only has an 8.72% gap from the optimum solved by the LINGO solver.

cs.NI↗

arXiv2026

2.4 GHz Flip-flop Device within Nonequilibrium Superconducting Diode

Superconducting diode effect exhibits asymmetric critical supercurrent and has profound implications for condensed matter physics. The technical appeals of such superconducting diodes are their ultrahigh on-off ratio and diode efficiency for superconducting electronics owing to the dissipationless supercurrent therein. However, realizing superconducting diode operation at high working frequency, which is a key requirement for practical applications, remains elusive and challenging. Here, we demonstrate a polarity-controllable superconducting diode with non-equilibrium Josephson junction and its edge-triggered flip-flop operation at a high frequency up to 2.4 GHz, within a van der Waals superconductor 2M-WS$_2$. By simply tuning the thickness of superconducting 2M-WS$_2$ nanoflakes to engineer inversion asymmetry in the junction, we achieve a high diode efficiency of 67% and an on-off ratio exceeding 10$^5$. Importantly, the pulse width and duty cycle of output pulse signals in such superconducting diode flip-flop devices can be controlled in a broadband frequency range crossing 12 orders of magnitude. Theoretical analysis reveals that the non-equilibrium dynamic nature of supercurrent in these Josephson junctions enables such a high diode operating frequency and the polarity control of supercurrent. The 2.4 GHz non-equilibrium Josephson diode developed here provides a promising platform for advanced superconducting logic circuits and broadband telecommunication applications.

cond-mat.supr-con↗

arXiv2026

Satellite NOMA for Direct-to-Cell Communications: Fundamentals, Protocols, and Opportunities

Direct-to-cell (DTC) satellite communication is regarded as one of the most recent technologies that provides global connectivity. However, with the growing number of wireless users and devices, the design of DTC communications must satisfy the requirements of high-scale capabilities and efficient spectrum utilization. To this end, integrating satellite communications with advanced multiple-access techniques, such as non-orthogonal multiple access (NOMA), has attracted considerable interest in developing NOMA-DTC communications. In this article, we first introduce the fundamentals of NOMA-DTC communications, including architectural fundamentals, system design aspects, and potential applications. Given the various cooperative modes and the still-evolving satellite network (SatNet) architectures, such as cooperative SatNets and multi-tier SatNets, we explore protocols that suit future SatNets and enhance system performance. Furthermore, a case study is conducted to investigate the benefits of NOMA schemes for DTC communications and to compare them with OMA schemes. Finally, to inspire further research, several opportunities for NOMA-DTC communications are presented.

eess.SP↗

arXiv2026

SF-AMS: Strategic Forgetting for Structured Memory in LLM Agent

Managing long-context dependencies remains a primary bottleneck in LLM agents, as redundant and irrelevant information can degrade multi-step reasoning. Strategic Forgetting for Agent Memory Systems (SF-AMS) is proposed as a framework for maintaining compact high-utility memory by modeling the long-term importance of memory units. SF-AMS replaces static retrieval and heuristic decay with a utility-driven survival mechanism that updates memory importance from usage redundancy and temporal signals, inducing a hierarchical memory structure that prioritizes stable entity-consistent information while filtering noise. On top of this, Composite Importance Scoring integrates semantic and entity level signals to improve retrieval robustness. Experiments on LoCoMo and LongMemEval-s show consistent gains over strong state of the art baselines including LightMem MemO and A-Mem. The largest improvement appears in multi-hop reasoning under Qwen2.5-7B where SF-AMS achieves plus 9.65 F1 over the strongest baseline followed by temporal reasoning under GPT-4o-mini plus 6.91 F1 and open-domain tasks plus 6.53 F1 demonstrating strong cross backbone generalization. These results show that modeling memory importance as a dynamic utility signal is critical for reliable long-context reasoning.

cs.AI↗

arXiv2026

GeoWeaver: Grounding Visual Tokens with Geometric Evidence before Scene Reasoning

Spatio-temporal reasoning in vision-language models requires visual representations that preserve physical geometry rather than merely semantic appearance. Recent multimodal models incorporate geometric information through structural branches, 3D-aware supervision, reasoning-stage fusion, or long-horizon memory. While these approaches demonstrate the importance of geometry for spatial intelligence, they typically treat geometric cues as a shared signal across all visual tokens. We note that this overlooks a finer-grained challenge: different visual tokens require different geometric evidence depending on their spatial roles. To address this limitation, we introduce GeoWeaver, a pre-reasoning geometric grounding framework that treats geometry as a representational prerequisite for spatio-temporal reasoning. GeoWeaver constructs a multi-level geometry bank from a frozen geometry encoder and performs token-adaptive geometric evidence allocation, enabling each visual token to retrieve the most relevant geometric abstractions. The selected evidence is incorporated into visual tokens via a residual grounding operation prior to language modeling, yielding geometry-grounded representations for downstream reasoning. Extensive evaluations on spatial reasoning benchmarks demonstrate that GeoWeaver consistently enhances geometry-aware reasoning while retaining general multimodal capabilities. This indicates that geometric information yields the greatest benefit not as a late-fusion auxiliary signal but as a fundamental prerequisite that shapes the representational foundation on which large language models perform reasoning. All source code and models will be released at https://github.com/yahooo-m/GeoWeaver .

cs.CV↗

arXiv2026

Offline Policy Optimization with Posterior Sampling

A fundamental challenge in model-based offline reinforcement learning (RL) lies in the trade-off between generalization and robustness against exploitation errors in out-of-distribution (OOD) regions. While OOD samples may capture valid underlying physical dynamics, they also introduce the risk of model exploitation. Existing methods typically address this risk through excessive pessimistic regularization, which ensures robustness but often sacrifices generalization. To overcome this limitation, we propose Posterior Sampling-based Policy Optimization (PSPO), which formulates dynamics modeling as a Bayesian inference process to derive a posterior that explicitly quantifies model fidelity. Through the integration of posterior sampling and constrained policy optimization, our method leverages dynamics-consistent OOD transitions for generalization while ensuring robustness against model exploitation. Theoretically, we formulate Q-value estimation under posterior sampling as a stochastic approximation problem and establish its convergence. We decompose policy optimization into a sequence of constrained subproblems, demonstrating that solving these subproblems guarantees monotonic improvement until convergence. Experiments on standard benchmarks validate that PSPO achieves superior performance compared to state-of-the-art baselines.

cs.AI↗

arXiv2026

Report of the 5th PVUW Challenge: Towards More Diverse Modalities in Pixel-Level Understanding

This report summarizes the objectives, datasets, and top-performing methodologies of the 2026 Pixel-level Video Understanding in the Wild (PVUW) Challenge, hosted at CVPR 2026, which evaluates state-of-the-art models under highly unconstrained conditions. To provide a comprehensive assessment, the 2026 edition features three specialized tracks: the MOSE track for tracking objects within densely cluttered and severely occluded scenarios; the MeViS-Text track for localizing targets via motion-focused linguistic expressions; and the newly inaugurated MeViS-Audio track, which pioneers acoustic-driven object segmentation. By introducing previously unreleased challenging data and analyzing the cutting-edge, multimodal solutions submitted by participants, this report highlights the community's latest technical advancements and charts promising future directions for robust video scene comprehension.

cs.CV↗

arXiv2026

APRVOS: 1st Place Winner of 5th PVUW MeViS-Audio Track

This report presents an Audio-aware Referring Video Object Segmentation (Ref-VOS) pipeline tailored to the MEVIS\_Audio setting, where the referring expression is provided in spoken form rather than as clean text. Compared with a standard Sa2VA-based Ref-VOS pipeline, the proposed system introduces two additional front-end stages: speech transcription and visual existence verification. Specifically, we first employ VibeVoice-ASR to convert long-form spoken input into a structured textual transcript. Since audio-derived queries are inherently noisy and may describe entities that are not visually present in the video, we then introduce an Omni-based judgment module to determine whether the transcribed target can be grounded in the visual content. If the target is judged to be absent, the pipeline terminates early and outputs all-zero masks. Otherwise, the transcript is transformed into a segmentation-oriented prompt and fed into Sa2VA to obtain a coarse mask trajectory over the full video. Importantly, this trajectory is treated as an initial semantic hypothesis rather than a final prediction. On top of it, an agentic refinement layer evaluates query reliability, temporal relevance, anchor quality, and potential error sources, and may invoke SAM3 to improve spatial boundary precision and temporal consistency. The resulting framework explicitly decomposes the MEVIS\_Audio task into audio-to-text conversion, visual existence verification, coarse video segmentation, and agent-guided refinement. Such a staged design is substantially more appropriate for audio-conditioned Ref-VOS than directly sending noisy ASR outputs into a segmentation model.

cs.SD↗

arXiv2026

Category-based and Popularity-guided Video Game Recommendation: A Balance-oriented Framework

In recent years, the video game industry has experienced substantial growth, presenting players with a vast array of game choices. This surge in options has spurred the need for a specialized recommender system tailored for video games. However, current video game recommendation approaches tend to prioritize accuracy over diversity, potentially leading to unvaried game suggestions. In addition, the existing game recommendation methods commonly lack the ability to establish strict connections between games to enhance accuracy. Furthermore, many existing diversity-focused methods fail to leverage crucial item information, such as item category and popularity during neighbor modeling and message propagation. To address these challenges, we introduce a novel framework, called CPGRec, comprising three modules, namely accuracy-driven, diversity-driven, and comprehensive modules. The first module extends the state-of-the-art accuracy-focused game recommendation method by connecting games in a more stringent manner to enhance recommendation accuracy. The second module connects neighbors with diverse categories within the proposed game graph and harnesses the advantages of popular game nodes to amplify the influence of long-tail games within the player-game bipartite graph, thereby enriching recommendation diversity. The third module combines the above two modules and employs a new negative-sample rating score reweighting method to balance accuracy and diversity. Experimental results on the Steam dataset demonstrate the effectiveness of our proposed method in improving game recommendations. The dataset and source codes are anonymously released at: https://github.com/CPGRec2024/CPGRec.git.

cs.IR↗

arXiv2026

CPGRec+: A Balance-oriented Framework for Personalized Video Game Recommendations

The rapid expansion of gaming industry requires advanced recommender systems tailored to its dynamic landscape. Existing Graph Neural Network (GNN)-based methods primarily prioritize accuracy over diversity, overlooking their inherent trade-off. To address this, we previously proposed CPGRec, a balance-oriented gaming recommender system. However, CPGRec fails to account for critical disparities in player-game interactions, which carry varying significance in reflecting players' personal preferences and may exacerbate over-smoothness issues inherent in GNN-based models. Moreover, existing approaches underutilize the reasoning capabilities and extensive knowledge of large language models (LLMs) in addressing these limitations. To bridge this gap, we propose two new modules. First, Preference-informed Edge Reweighting (PER) module assigns signed edge weights to qualitatively distinguish significant player interests and disinterests while then quantitatively measuring preference strength to mitigate over-smoothing in graph convolutions. Second, Preference-informed Representation Generation (PRG) module leverages LLMs to generate contextualized descriptions of games and players by reasoning personal preferences from comparing global and personal interests, thereby refining representations of players and games. Experiments on \textcolor{black}{two Steam datasets} demonstrate CPGRec+'s superior accuracy and diversity over state-of-the-art models. The code is accessible at https://github.com/HsipingLi/CPGRec-Plus.

cs.IR↗

arXiv2026

LLM-Driven Large-Scale Spectrum Access

Efficient spectrum management in massive-scale wireless networks is increasingly challenged by explosive action spaces and the computational intractability of traditional optimization. This study proposes a LLM-Driven Large-Scale Spectrum Access (LSA) framework rooted in Group Relative Policy Optimization (GRPO). To overcome the computational intractability caused by ultra-long prompts in large-scale scenarios, we develop a hierarchical state serialization mechanism that synthesizes global environment statistics with localized critical constraints, enabling the LLM to perform high-dimensional reasoning within a bounded context window. Simulation results under strictly time-bounded inference protocols reveal that the code-driven paradigm eliminates the Supervised Fine-Tuning (SFT) cold-start bottleneck and leverages direct execution feedback to achieve superior scaling laws. The framework maintains robust spectral utility and generalization across varying network scales, yielding consistent and empirically superior performance over stochastic heuristics, and surpassing partitioned classical solvers in ultra-dense regimes under matched compute budgets. Code is available at https://github.com/Xtdzs/LLM-Driven-Large-Scale-Spectrum-Access.

cs.NI↗

arXiv2026

Multi-Turn Reasoning LLMs for Task Offloading in Mobile Edge Computing

Emerging computation-intensive applications impose stringent latency requirements on resource-constrained mobile devices. Mobile Edge Computing (MEC) addresses this challenge through task offloading. However, designing effective policies remains difficult due to dynamic task arrivals, time-varying channels, and the spatio-temporal coupling of server queues. Conventional heuristics lack adaptability, while Deep Reinforcement Learning (DRL) suffers from limited generalization and architectural rigidity, requiring retraining when network topology changes. Although Large Language Models (LLMs) offer semantic reasoning capabilities, standard Supervised Fine-Tuning (SFT) yields myopic policies that greedily minimize immediate latency without accounting for long-term system evolution. To address these limitations, we propose COMLLM, a generative framework that enables foresighted decision-making in MEC systems. COMLLM integrates Group Relative Policy Optimization (GRPO) with a Look-Ahead Collaborative Simulation (LACS) mechanism, which performs multi-step Monte Carlo rollouts while jointly modeling server queue dynamics. By incorporating these rollouts into the reward design, the framework captures the long-term impact of current decisions on future system states. Experimental results demonstrate that COMLLM achieves near-optimal latency and improved load-balancing fairness. Notably, it exhibits zero-shot topological scalability, allowing a model trained on small-scale networks to generalize to larger, unseen topologies without retraining, outperforming SFT, DRL, and heuristic baselines.

cs.LG↗

arXiv2026

PaveBench: A Versatile Benchmark for Pavement Distress Perception and Interactive Vision-Language Analysis

Pavement condition assessment is essential for road safety and maintenance. Existing research has made significant progress. However, most studies focus on conventional computer vision tasks such as classification, detection, and segmentation. In real-world applications, pavement inspection requires more than visual recognition. It also requires quantitative analysis, explanation, and interactive decision support. Current datasets are limited. They focus on unimodal perception. They lack support for multi-turn interaction and fact-grounded reasoning. They also do not connect perception with vision-language analysis. To address these limitations, we introduce PaveBench, a large-scale benchmark for pavement distress perception and interactive vision-language analysis on real-world highway inspection images. PaveBench supports four core tasks: classification, object detection, semantic segmentation, and vision-language question answering. It provides unified task definitions and evaluation protocols. On the visual side, PaveBench provides large-scale annotations and includes a curated hard-distractor subset for robustness evaluation. It contains a large collection of real-world pavement images. On the multimodal side, we introduce PaveVQA, a real-image question answering (QA) dataset that supports single-turn, multi-turn, and expert-corrected interactions. It covers recognition, localization, quantitative estimation, and maintenance reasoning. We evaluate several state-of-the-art methods and provide a detailed analysis. We also present a simple and effective agent-augmented visual question answering framework that integrates domain-specific models as tools alongside vision-language models. The dataset is available at: https://huggingface.co/datasets/MML-Group/PaveBench.

cs.CV↗

arXiv2026

LinearARD: Linear-Memory Attention Distillation for RoPE Restoration

The extension of context windows in Large Language Models is typically facilitated by scaling positional encodings followed by lightweight Continual Pre-Training (CPT). While effective for processing long sequences, this paradigm often disrupts original model capabilities, leading to performance degradation on standard short-text benchmarks. We propose LinearARD, a self-distillation method that restores Rotary Position Embeddings (RoPE)-scaled students through attention-structure consistency with a frozen native-RoPE teacher. Rather than matching opaque hidden states, LinearARD aligns the row-wise distributions of dense $Q/Q$, $K/K$, and $V/V$ self-relation matrices to directly supervise attention dynamics. To overcome the quadratic memory bottleneck of $n \times n$ relation maps, we introduce a linear-memory kernel. This kernel leverages per-token log-sum-exp statistics and fuses logit recomputation into the backward pass to compute exact Kullback-Leibler divergence and gradients. On LLaMA2-7B extended from 4K to 32K, LinearARD recovers 98.3\% of the short-text performance of state-of-the-art baselines while surpassing them on long-context benchmarks. Notably, our method achieves these results using only \textbf{4.25M} training tokens compared to the \textbf{256M} tokens required by LongReD and CPT. Our code is available at https://github.com/gracefulning/LinearARD.

cs.CL↗