arXiv subjects

Huixiang Zhang

Publications and source records attributed to Huixiang Zhang.

6 recordsLinked to original sources

arXiv2026

Do Latent Channels Actually Communicate? A Causal Audit of Latent Multi-Agent LLM

Latent communication in large language model (LLM)-based multi-agent systems (MAS) transmits continuous internal representations instead of text, but greater representational capacity does not establish that the receiver uses task-relevant information. End-task performance alone also cannot reveal whether an observed effect depends on message presence, content generated for the evaluated example, or information supplied by a separate agent. We introduce a causal audit that applies controlled message replacements at the boundary where the sender-produced representation enters the receiver. Four message settings support five measurements of encoded sender information, receiver sensitivity to message presence and identity, the task value of example-specific content, and the additional value supplied by a separate agent. We apply the audit to latent relay with Qwen3-4B and Qwen3-8B on GSM8K, ARC-C, and MATH-500. On GSM8K, the Qwen3-4B overall performance effect of -1.00 percentage point decomposes into a -6.17-point effect retained by an other-example message and a +5.17-point effect attributable to example-specific content; both component directions reverse at 8B. On MATH-500, the Qwen3-4B gain of 15.00 points comprises 8.33 points retained by an other-example message and 6.67 points attributable to example-specific content, while the 8B gain is dominated by the former component. Self-substitution comparisons further show that example-specific content and other-agent value are distinct. These results show that aggregate accuracy does not identify how a latent message affects the receiver and motivate controlled message comparisons as a standard evaluation for latent communication.

cs.AI↗

arXiv2026

Q-GARS: Quantum-inspired Robust Microservice Chaining Scheduling

Microservice-based applications are characterized by stochastic latencies arising from long-tail execution patterns and heterogeneous resource constraints across computational nodes. To address this challenge, we first formulate the problem using Quadratic Unconstrained Binary Optimization (QUBO), which aligns the problem with emerging quantum-optimization paradigms. Building upon this, we propose Q-GARS (Quantum-Guided Adaptive Robust Scheduling), a hybrid framework that integrates the QUBO model with Simulated Quantum Annealing (SQA) based combinatorial search and online rescheduling mechanisms, enabling global microservice rank generation and real-time robust adjustment. We treat the SQA-produced rank as a soft prior, and update a closed-loop trust weight to adaptively switch and mix between this prior and a robust proportional-fairness allocator, maintaining robustness under prediction failures and runtime disturbances. Simulation results demonstrate that Q-GARS achieves an average weighted completion time improvement of 2.1\% relative to a greedy baseline of the remaining shortest processing-time (SRPT), with performance gains reaching up to 16.8\% in heavy-tailed latency. The adaptive mechanism reduces tail latency under high-variance conditions. In addition, Q-GARS achieves a mean node resource utilization rate of 0.817, which is 1.1 percentage points above the robust baseline (0.806).

cs.SE↗

arXiv2026

Quantum Takes Flight: Two-Stage Resilient Topology Optimization for UAV Networks

Next-generation Unmanned Aerial Vehicle (UAV) communication networks must maintain reliable connectivity under rapid topology changes, fluctuating link quality, and time-critical data exchange. Existing topology control methods rely on global optimization to produce a single optimal topology or involve high computational complexity, which limits adaptability in dynamic environments. This paper presents a two-stage quantum-assisted framework for efficient and resilient topology control in dynamic UAV networks by exploiting quantum parallelism to generate a set of high-quality and structurally diverse candidate topologies. In the offline stage, we formulate the problem as a Quadratic Unconstrained Binary Optimization (QUBO) model and leverage quantum annealing (QA) to parallelly sample multiple high-quality and structurally distinct topologies, providing a rich solution space for adaptive decision-making. In the online stage, a lightweight classical selection mechanism rapidly identifies the most suitable topology based on real-time link stability and channel conditions, substantially reducing the computation delay. The simulation results show that, compared to a single static optimal topology, the proposed framework improves performance retention by 6.6% in a 30-second dynamic window. Moreover, relative to the classic method, QA achieves an additional 5.15% reduction in objective value and a 28.3% increase in solution diversity. These findings demonstrate the potential of QA to enable fast and robust topology control for next-generation UAV communication networks.

cs.NI↗

arXiv2025

Quantum-Guided Test Case Minimization for LLM-Based Code Generation

Precisely controlling Large Language Models (LLMs) to generate efficient and concise code is a central challenge in software engineering. We introduce a framework based on Test-Driven Development (TDD) that transforms code specification into a combinatorial optimization task. The framework first prompts an LLM to generate a test suite, then formulates the Test Case Minimization (TCM) problem as a Quadratic Unconstrained Binary Optimization (QUBO) model. This QUBO paradigm is compatible with both classical solvers and emerging hardware such as quantum annealers. Experimentally, quantum annealing solves the core TCM task 16 times faster than simulated annealing. This performance underpins our end-to-end framework, which reduces total token consumption by 36.5\% and significantly improves code quality. This work demonstrates a powerful synergy between generative AI and combinatorial optimization in software engineering, highlighting the critical importance of precise model formulation.

cs.SE↗

arXiv2025

Coherent Optical Quantum Computing-Aided Resource Optimization for Transportation Digital Twin Construction

Constructing realistic digital twins for applications such as training autonomous driving models requires the efficient allocation of real-world data, yet data sovereignty regulations present a major challenge. To address this, we tackle the optimization problem faced by metaverse service providers (MSPs) responsible for allocating geographically constrained data resources. We propose a two-stage stochastic integer programming (SIP) model that incorporates reservation and on-demand planning, enabling MSPs to efficiently subscribe and allocate data from specific regions to clients for training their models on local road conditions. The SIP model is transformed into a quadratic unconstrained binary optimization (QUBO) formulation and implemented for the first time at a practical scale on a 550-qubit coherent Ising machine (CIM), representing an exploratory step toward future quantum computing paradigms. Our approach introduces an MSP-centric framework for compliant data collection under sovereignty constraints, a hybrid cost model combining deterministic fees with probabilistic penalties, and a practical implementation on quantum hardware. Experimental results demonstrate that CIM-based optimization finds high-quality solutions with millisecond-scale ($10^3$ second) computation times, significantly outperforming quantum-inspired solvers like PyQUBO. Although classical solvers such as Gurobi can achieve marginally better solution quality, CIM is orders of magnitude faster, establishing a practical paradigm for quantum-enhanced resource management.

cs.OH↗

arXiv2025

LLM-QUBO: An End-to-End Framework for Automated QUBO Transformation from Natural Language Problem Descriptions

Quantum annealing offers a promising paradigm for solving NP-hard combinatorial optimization problems, but its practical application is severely hindered by two challenges: the complex, manual process of translating problem descriptions into the requisite Quadratic Unconstrained Binary Optimization (QUBO) format and the scalability limitations of current quantum hardware. To address these obstacles, we propose a novel end-to-end framework, LLM-QUBO, that automates this entire formulation-to-solution pipeline. Our system leverages a Large Language Model (LLM) to parse natural language, automatically generating a structured mathematical representation. To overcome hardware limitations, we integrate a hybrid quantum-classical Benders' decomposition method. This approach partitions the problem, compiling the combinatorial complex master problem into a compact QUBO format, while delegating linearly structured sub-problems to classical solvers. The correctness of the generated QUBO and the scalability of the hybrid approach are validated using classical solvers, establishing a robust performance baseline and demonstrating the framework's readiness for quantum hardware. Our primary contribution is a synergistic computing paradigm that bridges classical AI and quantum computing, addressing key challenges in the practical application of optimization problem. This automated workflow significantly reduces the barrier to entry, providing a viable pathway to transform quantum devices into accessible accelerators for large-scale, real-world optimization challenges.

cs.LG↗