arXiv subjects

Xi Yu

Publications and source records attributed to Xi Yu.

At least 19 recordsLinked to original sources

arXiv2026

Decentralized Federated Learning for Heterogeneous Multi-Task Semantic Communication

Collaborative training in distributed semantic communication (DSC) networks typically relies on decentralized federated learning (DFL). However, pushing topology-agnostic aggregation into heterogeneous, multi-task environments creates a fundamental bottleneck: it drives negative transfer and overconsensus bias (OCB). This paper introduces a personalized DSC framework that cuts off this cross-task interference. At the node level, a policy-driven multi-path routing mechanism separates task-specific features from shared representations to preserve local fidelity. Across the network, we deploy a "communicationwhile- aggregation" protocol. It calibrates a column-stochastic consensus matrix using task affinities. This limits the system to absorbing complementary knowledge while actively blocking mismatched parameter updates. To bound the convergence, we derive a unified Lyapunov drift analysis. We reveal a strict Ushaped trade-off: deeper topological mixing reduces variance but amplifies structural OCB. Resolving this tension yields a closed-form expression for the optimal aggregation depth. We evaluate the proposed framework on NYU-v2, where the results reveal a clear trade-off between insufficient aggregation and excessive topological mixing. At the analytically derived optimal aggregation depth, our method achieves a 4.77% global relative improvement over the no-aggregation baseline and outperforms decentralized FedAvg, FedAMP, and heuristic max aggregation. We further evaluate the framework on Taskonomy and imperfect wireless links to examine the effects of network-size variation and wireless-link reliability.

cs.AI↗

arXiv2026

Inverse mask design for interference lithography using automatic differentiable wave propagation

Interference lithography (IL) is powerful for fabricating high-resolution periodic nanostructures, but designing masks to produce non-periodic patterns remains challenging. We introduce a gradient-based optimization framework for binary IL mask design using automatic differentiation. The forward model is implemented using the differentiable angular spectrum method (ASM). The inverse mask design is formulated as an optimization problem, where the mask logits are updated through backpropagation of the loss between the simulated field amplitude and the target pattern. We optimize a mask that reproduces a target pattern with only 0.1% isolated pixel-level defects, resolving features at half the mask pixel pitch. To scale mask optimization, we employ the shifted ASM, which partitions the mask into patches that are propagated independently and summed at the image plane. For a 3.84 mm$\times$3.84 mm mask, shifted ASM with 16 patches reduces peak GPU memory by 3.8$\times$ at only 1.3$\times$ runtime cost relative to standard ASM. With gradient checkpointing, peak memory is reduced by 7.4$\times$ at 2$\times$ runtime. Distributing across multiple GPUs further accelerates the optimization. This work establishes a physics-informed, machine learning-driven approach for IL mask design, moving a step further towards complex, non-periodic patterns. The source code is available at https://github.com/chuntian236/holography-optimization.git .

physics.optics↗

arXiv2026

Adaptive Dual-Path Framework for Covert Semantic Communication

This paper proposes a novel adaptive dual-path framework for covert semantic communication (SemCom), which integrates covert information transmission with task-oriented semantic coding. Unlike conventional covert communication methods that embed hidden messages through power-domain signal superposition, our framework embeds covert data within task-specific features via semantic-level intrinsic encoding. This new architecture introduces dual encoding paths with adaptive block selection: an Explicit path for public task execution and a Stego path that jointly encodes both public and covert information through contrastive representation alignment. A Gumbel-Softmax enabled adaptive path selection mechanism dynamically activates network blocks based on task require- ments. We formulate a multi-objective optimization framework that simultaneously ensures accurate semantic understanding and reliable covert transmission. We rigorously evaluate our framework's security against a powerful, independently trained attacker. Experimental results on the Cityscapes dataset demon- strate a state-of-the-art level of covertness: our method suppresses the attacker's detection accuracy to a near-random guessing level of 56.12%. This robust security is achieved while simultaneously maintaining superior performance on the primary semantic tasks compared to the baselines.

cs.AI↗

arXiv2026

Hierarchy-Guided Multimodal Representation Learning for Taxonomic Inference

Accurate biodiversity identification from large-scale field data is a foundational problem with direct impact on ecology, conservation, and environmental monitoring. In practice, the core task is taxonomic prediction - inferring order, family, genus, or species from imperfect inputs such as specimen images, DNA barcodes, or both. Existing multimodal methods often treat taxonomy as a flat label space and therefore fail to encode the hierarchical structure of biological classification, which is critical for robustness under noise and missing modalities. We present two end-to-end variants for hierarchy-aware multimodal learning: CLiBD-HiR, which introduces Hierarchical Information Regularization (HiR) to shape embedding geometry across taxonomic levels, yielding structured and noise-robust representations; and CLiBD-HiR-Fuse, which additionally trains a lightweight fusion predictor that supports image-only, DNA-only, or joint inference and is resilient to modality corruption. Across large-scale biodiversity benchmarks, our approach improves taxonomic classification accuracy by over 14 percent compared to strong multimodal baselines, with particularly large gains under partial and corrupted DNA conditions. These results highlight that explicitly encoding biological hierarchy, together with flexible fusion, is key for practical biodiversity foundation models.

cs.CV↗

arXiv2026

Joint Routing and Model Pruning for Decentralized Federated Learning in Bandwidth-Constrained Multi-Hop Wireless Networks

Decentralized federated learning (D-FL) enables privacy-preserving training without a central server, but multi-hop model exchanges and aggregation are often bottlenecked by communication resource constraints. To address this issue, we propose a joint routing-and-pruning framework that optimizes routing paths and pruning rates to maintain communication latency within prescribed limits. We analyze how the sum of model biases across all clients affects the convergence bound of D-FL and formulate an optimization problem that maximizes the model retention rate to minimize these biases under communication constraints. Further analysis reveals that each client's model retention rate is path-dependent, which reduces the original problem to a routing optimization. Leveraging this insight, we develop a routing algorithm that selects latency-efficient transmission paths, allowing more parameters to be delivered within the time budget and thereby improving D-FL convergence. Simulations demonstrate that, compared with unpruned systems, the proposed framework reduces average transmission latency by 27.8% and improves testing accuracy by approximately 12%. Furthermore, relative to standard benchmark routing algorithms, the proposed routing method improves accuracy by roughly 8%.

cs.LG↗

arXiv2026

Hybrid topology control: a dynamic leader-based distributed edge-addition and deletion mechanism

Coordinated operations of multi-robot systems (MRS) require agents to maintain communication connections to accomplish team objectives. However, maintaining the connections imposes costs in terms of restricted robot mobility, resulting in suboptimal team performance. In this work, we consider a realistic MRS framework in which agents are subject to unknown dynamical disturbances and experience communication delays. Most existing works on connectivity maintenance use consensus-based frameworks for graph reconfiguration, where decision-making time scales with the number of nodes and requires multiple rounds of communication, making them ineffective under communication delays. To address this, we propose a novel leader-based decision-making algorithm that uses a central node for efficient real-time reconfiguration, reducing decision-making time to depend on the graph diameter rather than the number of nodes and requiring only one round of information transfer through the network. We propose a novel method for estimating robot locations within the MRS that actively accounts for unknown disturbances and the communication delays. Using these position estimates, the central node selects a set of edges to delete while allowing the formation of new edges, aiming to keep the diameter of the new graph within a threshold. We provide numerous simulation results to showcase the efficacy of the proposed method.

math.OC↗

arXiv2026

AutoSizer: Automatic Sizing of Analog and Mixed-Signal Circuits via Large Language Model (LLM) Agents

The design of Analog and Mixed-Signal (AMS) integrated circuits remains heavily reliant on expert knowledge, with transistor sizing a major bottleneck due to nonlinear behavior, high-dimensional design spaces, and strict performance constraints. Existing Electronic Design Automation (EDA) methods typically frame sizing as static black-box optimization, resulting in inefficient and less robust solutions. Although Large Language Models (LLMs) exhibit strong reasoning abilities, they are not suited for precise numerical optimization in AMS sizing. To address this gap, we propose AutoSizer, a reflective LLM-driven meta-optimization framework that unifies circuit understanding, adaptive search-space construction, and optimization orchestration in a closed loop. It employs a two-loop optimization framework, with an inner loop for circuit sizing and an outer loop that analyzes optimization dynamics and constraints to iteratively refine the search space from simulation feedback. We further introduce AMS-SizingBench, an open benchmark comprising 24 diverse AMS circuits in SKY130 CMOS technology, designed to evaluate adaptive optimization policies under realistic simulator-based constraints. AutoSizer experimentally achieves higher solution quality, faster convergence, and higher success rate across varying circuit difficulties, outperforming both traditional optimization methods and existing LLM-based agents.

cs.AI↗

arXiv2025

Training Report of TeleChat3-MoE

TeleChat3-MoE is the latest series of TeleChat large language models, featuring a Mixture-of-Experts (MoE) architecture with parameter counts ranging from 105 billion to over one trillion,trained end-to-end on Ascend NPU cluster. This technical report mainly presents the underlying training infrastructure that enables reliable and efficient scaling to frontier model sizes. We detail systematic methodologies for operator-level and end-to-end numerical accuracy verification, ensuring consistency across hardware platforms and distributed parallelism strategies. Furthermore, we introduce a suite of performance optimizations, including interleaved pipeline scheduling, attention-aware data scheduling for long-sequence training,hierarchical and overlapped communication for expert parallelism, and DVM-based operator fusion. A systematic parallelization framework, leveraging analytical estimation and integer linear programming, is also proposed to optimize multi-dimensional parallelism configurations. Additionally, we present methodological approaches to cluster-level optimizations, addressing host- and device-bound bottlenecks during large-scale training tasks. These infrastructure advancements yield significant throughput improvements and near-linear scaling on clusters comprising thousands of devices, providing a robust foundation for large-scale language model development on hardware ecosystems.

cs.CL↗

arXiv2025

Synchronous Differential Hot-charge Emission Spectroscopy: The Principle

Energy-level alignment (ELA) at buried interfaces between electrode and molecular materials sets charge injection barriers, carrier selectivity, and ultimately device efficiency, yet it is challenging to quantify under operating conditions. Hot-charge emission spectroscopy (HotES) probes ELA by injecting ballistic carriers across a tunneling oxide. Yet, the technique inherently convolutes the molecular response with a strong, energy-dependent tunneling background, complicating the isolation of the true ELA. We introduce synchronous differential HotES (sd-HotES), defined as the ratio of the differential conductance of the hot-charge and tunneling channels of the HotES. Physical modeling and numerical simulations validate that this ratio directly reconstructs the intrinsic molecular charge transmission, enabling the threshold-free and probe-bias-insensitive extraction of ELA. By effectively eliminating the masking tunneling background, sd-HotES substantially boosts detection sensitivity; weak spectral features previously hidden in conventional HotES become clearly resolvable, as demonstrated in lock-in simulations including realistic noise. This study establishes the fundamental operating principles of sd-HotES and highlights it as a powerful, broadly applicable strategy for accessing buried interface properties for the study of molecular and hybrid devices.

physics.chem-ph↗

arXiv2025

A*-based Temporal Logic Path Planning with User Preferences on Relaxed Task Satisfaction

In this work, we consider the problem of planning for temporal logic tasks in large robot environments. When full task compliance is unattainable, we aim to achieve the best possible task satisfaction by integrating user preferences for relaxation into the planning process. Utilizing the automata-based representations for temporal logic goals and user preferences, we propose an A*-based planning framework. This approach effectively tackles large-scale problems while generating near-optimal high-level trajectories. To facilitate this, we propose a simple, efficient heuristic that allows for planning over large robot environments in a fraction of time and search memory as compared to uninformed search algorithms. We present extensive case studies to demonstrate the scalability, runtime analysis as well as empirical bounds on the suboptimality of the proposed heuristic.

cs.RO↗

arXiv2025

Maximizing the nondemolition nature of a quantum measurement via an adaptive readout protocol

Quantum error correction (QEC) requires non-invasive measurements for fault tolerant quantum computing. Deviations from ideal quantum non-demolition (QND) measurements can disturb the encoded information. To address this challenge, we develop a readout protocol for a $D-$dimensional system that, after a single positive outcome, switches to probing only the $D{-}1$ remaining subspace. This adaptive switching strategy minimizes measurement-induced errors by relying on negative-result measurement results that do not perturb the Hamiltonian. We apply the protocol on an 8-dimensional $^{123}{\rm Sb}$ nuclear qudit in silicon, and achieve an increase in the readout fidelity from $(98.93\pm0.07)\%$ to $(99.61\pm0.04)\%$, while reducing threefold the overall readout time. To highlight the broader relevance of measurement-induced errors, we study a 10-dimensional $^{73}{\rm Ge}$ nuclear spin read out through Pauli spin blockade, revealing nuclear spin flips arising from hyperfine and quadrupole interactions. These results unveil the effect of non-ideal QND readout across diverse platforms, and introduce an efficient readout protocol that can be implemented with minimal FPGA logic on existing hardware.

quant-ph↗

arXiv2025

Automatic tuning of a donor in a silicon quantum device using machine learning

Donor spin qubits in silicon offer one- and two-qubit gates with fidelities beyond 99%, coherence times exceeding 30 seconds, and compatibility with industrial manufacturing methods. This motivates the development of large-scale quantum processors using this platform, and the ability to automatically tune and operate such complex devices. In this work, we present the first machine learning algorithm with the ability to automatically locate the charge transitions of an ion-implanted donor in a silicon device, tune single-shot charge readout, and identify the gate voltage parameters where tunnelling rates in and out the donor site are the same. The entire tuning pipeline is completed on the order of minutes. Our results enable both automatic characterisation and tuning of a donor in silicon devices faster than human experts.

cond-mat.mes-hall↗

arXiv2025

GenCellAgent: Generalizable, Training-Free Cellular Image Segmentation via Large Language Model Agents

Cellular image segmentation is essential for quantitative biology yet remains difficult due to heterogeneous modalities, morphological variability, and limited annotations. We present GenCellAgent, a training-free multi-agent framework that orchestrates specialist segmenters and generalist vision-language models via a planner-executor-evaluator loop (choose tool $\rightarrow$ run $\rightarrow$ quality-check) with long-term memory. The system (i) automatically routes images to the best tool, (ii) adapts on the fly using a few reference images when imaging conditions differ from what a tool expects, (iii) supports text-guided segmentation of organelles not covered by existing models, and (iv) commits expert edits to memory, enabling self-evolution and personalized workflows. Across seven cell-segmentation benchmarks spanning diverse microscopy modalities (4,718 images), this routing consistently matches or exceeds the best individual tool on every dataset and outperforms all baselines in overall accuracy. On out-of-distribution organelle data, GenCellAgent substantially outperforms specialist models that were not trained on the target domain, recovering structures that dedicated tools fail to detect. It also segments novel objects such as the Golgi apparatus via iterative text-guided refinement, with light human correction further boosting performance. Together, these capabilities provide a practical path to robust, adaptable cellular image segmentation without retraining, while reducing annotation burden and matching user preferences.

q-bio.QM↗

arXiv2025

Language-Native Materials Processing Design by Lightly Structured Text Database and Reasoning Large Language Model

Materials synthesis procedures are predominantly documented as narrative text in papers, protocols, and laboratory records, placing them beyond the reach of conventional data-driven optimization frameworks. This language-native character poses a particular challenge for complex, multistage processes such as the preparation of boron nitride nanosheets (BNNS), where outcomes depend on path-dependent choices in exfoliation, functionalization, and functionalization. Here, we recast synthesis planning of the materials as a text reasoning problem enabled by a lightly structured knowledge substrate that preserves the procedural logic and causal contexts while exposing computable elements for retrieval. Built on this representation, our framework combines semantic matching, lexical search, and parameter-aware filtering to support retrieval-augmented generation with more accurate and better-grounded synthesis guidance. We further introduce experience-augmented reasoning, in which iteratively refined text guides distilled from multi-source narratives support hypothesis generation, failure diagnosis, and protocol revision. We validated the framework in the targeted exfoliation of BNNS, a synthesis problem governed by multivariate constraints and limited transferability of literature protocols across laboratory settings. By integrating dispersed literature evidence with experimentally observed failure modes, the system converged within only three iterative rounds on a high-performing protocol that yielded high-quality ultrathin nanosheets meeting the target specifications, substantially shortening what is often a prolonged cycle of expert-led trial-and-error. By enabling language-native reasoning over procedural knowledge, this framework moves AI beyond literature assistance toward active synthesis planning, adaptation and acceleration in complex materials workflows.

cs.DB↗

arXiv2025

Specifying Composites in Structural Equation Modeling: Traditional, Recent and New Approaches

Composites, or linear combinations of variables, play an important role in multivariate behavioral research. They appear in the form of indices, inventories, formative constructs, parcels, and emergent variables. Although structural equation modeling is widely used to study relations between variables, current approaches to incorporating composites have one or more limitations. These limitations include not modeling composite creation, not employing weights as model parameters, not being able to estimate weights, not allowing for composites in endogenous model positions, and not being able to assess whether a composite fully transmits the effects of or on its components. To address these limitations, we propose two novel composite specifications. The first specification combines the refined H-O specification of composites with phantom variables and uses the inverse of the weights as free parameters. The second specification blends the H-O specification with the pseudo-indicator approach of Rose et al. and uses the weights of all but one indicator as free parameters. We demonstrate the applicability of these specifications using an empirical example. The paper concludes with recommendations for choosing among the available composite specifications, as well as suggestions for future research.

stat.ME↗

arXiv2025

A General Molecular-Scale Dynamic Memristor Model Based on Non-equilibrium Charge Transport Kinetics and Its Information Processing Capability in Reservoir Computing

Non-equilibrium molecular-scale dynamics, where fast electron transport couples with slow chemical state evolution, underpins the complex behaviors of molecular memristors, yet a general model linking these dynamics to neuromorphic computing remains elusive. We introduce a dynamic memristor model that integrates Landauer and Marcus electron transport theories with the kinetics of slow processes, such as proton/ion migration or conformational changes. This framework reproduces experimental conductance hysteresis and emulates synaptic functions like short-term plasticity (STP) and spike-timing-dependent plasticity (STDP). By incorporating the model into a reservoir computing (RC) architecture, we show that computational performance optimizes when input frequency and bias mapping range align with the molecular system's intrinsic kinetics. This chemistry-centric, bottom-up approach provides a theoretical foundation for molecular-scale neuromorphic computing, demonstrating how non-equilibrium molecular-scale dynamics can drive information processing in the post-Moore era.

physics.chem-ph↗

arXiv2025

Cooling Matters: Benchmarking Large Language Models and Vision-Language Models on Liquid-Cooled Versus Air-Cooled H100 GPU Systems

The unprecedented growth in artificial intelligence (AI) workloads, recently dominated by large language models (LLMs) and vision-language models (VLMs), has intensified power and cooling demands in data centers. This study benchmarks LLMs and VLMs on two HGX nodes, each with 8x NVIDIA H100 graphics processing units (GPUs), using liquid and air cooling. Leveraging GPU Burn, Weights and Biases, and IPMItool, we collect detailed thermal, power, and computation data. Results show that the liquid-cooled systems maintain GPU temperatures between 41-50 degrees Celsius, while the air-cooled counterparts fluctuate between 54-72 degrees Celsius under load. This thermal stability of liquid-cooled systems yields 17 percent higher performance (54 TFLOPs per GPU vs. 46 TFLOPs per GPU), improved performance per watt, reduced energy overhead, and greater system efficiency than the air-cooled counterparts. These findings underscore the energy and sustainability benefits of liquid cooling, offering a compelling path forward for hyperscale data centers s

cs.DC↗

arXiv2025

Enhancing Convergence, Privacy and Fairness for Wireless Personalized Federated Learning: Quantization-Assisted Min-Max Fair Scheduling

Personalized federated learning (PFL) offers a solution to balancing personalization and generalization by conducting federated learning (FL) to guide personalized learning (PL). Little attention has been given to wireless PFL (WPFL), where privacy concerns arise. Performance fairness of PL models is another challenge resulting from communication bottlenecks in WPFL. This paper exploits quantization errors to enhance the privacy of WPFL and proposes a novel quantization-assisted Gaussian differential privacy (DP) mechanism. We analyze the convergence upper bounds of individual PL models by considering the impact of the mechanism (i.e., quantization errors and Gaussian DP noises) and imperfect communication channels on the FL of WPFL. By minimizing the maximum of the bounds, we design an optimal transmission scheduling strategy that yields min-max fairness for WPFL with OFDMA interfaces. This is achieved by revealing the nested structure of this problem to decouple it into subproblems solved sequentially for the client selection, channel allocation, and power control, and for the learning rates and PL-FL weighting coefficients. Experiments validate our analysis and demonstrate that our approach substantially outperforms alternative scheduling strategies by 87.08%, 16.21%, and 38.37% in accuracy, the maximum test loss of participating clients, and fairness (Jain's index), respectively.

cs.DC↗