Searcharxiv⌕ Search

arXiv subjects

Yang Xu

Publications and source records attributed to Yang Xu.

At least 19 recordsLinked to original sources
arXiv2026

Beyond Callbacks: A Developer-friendly Programming Paradigm for Network Simulation

Network simulation plays a crucial role in both networking research and industry. Existing commonly-used Discrete Event Simulations (DES) are based on the callback mechanism for discrete events. However, as network functions grow in complexity, callbacks reveal a fundamental structural limitation: a callback can only represent a single point in a blocking operation and cannot capture the full operation as a coherent whole. This forces developers to manually decompose sequential network logic into scattered callback fragments, leading to stack ripping, callback hell, and heavy development and debugging burdens. We propose CoDES (\textbf{Co}routine-based \textbf{DES}), a coroutine-based programming paradigm for network simulation that addresses this structural limitation. CoDES targets development productivity rather than execution performance. Realizing CoDES in existing DES simulators is non-trivial, as coroutine suspension must be coordinated with the event scheduler, introducing challenges in coroutine resource conflicts, operation-level computation requirements, compatibility with existing callback-based simulators, and overhead. We implement a CoDES-based framework in ns-3, address all four challenges, and release it as open source. We evaluate CoDES by implementing Message Passing Interface (MPI), High Precision Congestion Control (HPCC), and Routing Information Protocol (RIP) in ns-3, spanning the transport, congestion control, and routing layers. Compared with callback-based counterparts, CoDES reduces lines of code by up to 51.4% and nesting depth by up to 50%, with no accuracy loss and negligible runtime and memory overhead.

cs.NI↗
arXiv2026

Online Bayesian Calibration under Gradual and Abrupt System Changes

Bayesian model calibration is central to digital twins and computer experiments, as it aligns model outputs with field observations by estimating calibration parameters and correcting systematic model bias. Classical Bayesian calibration introduces latent parameters and a discrepancy function to capture bias, but suffers from parameter-discrepancy confounding and is typically formulated as an offline procedure under a stationary data-generating assumption. These limitations are restrictive in modern digital twin applications, where systems evolve over time and may exhibit both gradual drift and abrupt regime shifts. While data assimilation methods enable sequential updates, they generally do not explicitly model systematic bias and are not robust to abrupt changes. We propose Bayesian Recursive Projected Calibration (BRPC), an online Bayesian calibration framework for streaming data under simulator mismatch and nonstationarity. BRPC extends projected calibration to the online setting through a two-stage filtering procedure: first, unconditional filtering of the calibration parameters, followed by conditional filtering of the model discrepancy function given the estimated calibration parameters, which mitigates parameter--discrepancy confounding, while the online filtering enables bias-aware adaptation under gradual system evolution. To handle abrupt changes, we integrate BRPC with restart mechanisms that detect regime shifts and reset the calibration process, yielding a unified approach to mixed nonstationarity. We provide theoretical analysis, including a tracking error bound of the model discrepancy estimate under gradual evolution, given accurate calibration parameter estimates and false-alarm and detection behavior of restart mechanisms. Empirical results on synthetic and plant-simulation benchmarks demonstrate BRPC outperforms the state-of-the-art approaches.

cs.LG↗
arXiv2026

Nonlinear kink modes for nonisentropic supersonic vortex sheets

In this paper, planar nonisentropic vortex sheets of the two-dimensional Euler equations under the stability condition are studied via nonlinear geometric optics, focusing on the interaction induced by a small-amplitude, high-frequency incident wave impinging on a planar vortex sheet. It is found that there exist three asymmetric critical incident angles at which resonance occurs, leading to a significant amplification of the reflected and refracted oscillatory waves. Moreover, by refining the asymptotic expansion, we deduce that the leading profiles of the highly oscillatory waves are described by initial-boundary value problems for Burgers transport equations, while the leading profile of the contact discontinuity front satisfies an initial value problem for a Hamilton--Jacobi-type equation, which reduces to a forced linear transport equation on one resonant branch. These results extend the resonant amplification mechanism from the isentropic setting to the nonisentropic Euler flow and reveal the asymmetric influence of entropy variations on the nonlinear evolution of compressible vortex sheets.

math.AP↗
arXiv2026

Automated Theorem Generation in First-Order Logic by Construction from Triangular Standard Contradictions

Automated theorem generation (ATG) seeks to construct new theorems rather than merely prove given conjectures. Existing approaches typically generate candidate statements that require subsequent proof or formal verification. We propose a constructive framework for ATG in first-order logic based on triangular standard contradictions. The method constructs minimally unsatisfiable clause sets from selected first-order literals, then forms theorems by retaining some clauses as premises and taking the negation of one clause, or a conjunction of clauses, as the conclusion. Minimal unsatisfiability ensures that the remaining premises are satisfiable, avoiding trivial inference by explosion, while unsatisfiability of the full set guarantees the resulting entailment. Theorems are therefore correct by construction, without a separate proof procedure for each instance. Extending the previously developed $Δ_1$ generator, we introduce seven methods, $Δ_2$--$Δ_8$, forming a unified family of eight theorem-generation methods. Four generators, $Δ_1$--$Δ_4$, have been implemented and released. We also formulate a general $Δ$-class method encompassing the family. Combinatorial analysis demonstrates the capacity to generate large numbers of distinct theorems systematically. This framework offers a principled foundation for automated theorem discovery and the construction of logically guaranteed knowledge.

cs.AI↗
arXiv2026

GUITestScape: Towards Open-set Evaluation on Exploratory GUI Testing

Exploratory GUI testing is a particularly demanding setting for MLLM agents: without predefined test scripts, an agent must autonomously navigate an application and discover defects through its own interaction. However, current evaluation falls short on two fronts. First, existing benchmarks focus almost exclusively on interaction defects, leaving display defects outside the evaluation frame. Second, evaluation protocols are bound to predefined defect annotations, collapsing the testing process into a single end-state judgment that conflates qualitatively distinct failure modes. To address these challenges, we present GUITestScape, an interactive benchmark covering 61 real-world Android applications and 508 preset defects spanning interaction and display types, and introduce GUIJudge, an open-set evaluator that decomposes an agent's testing trajectory into independently diagnosable capabilities. Experimental results demonstrate that GUIJudge achieves reliable process-aware evaluation beyond predefined annotations, substantially outperforming all baselines. Benchmarking on GUITestScape further reveals that detection remains the critical bottleneck for existing models across both defect types, and that integrating GUIJudge's verifiers into existing agents significantly boosts their detection performance without retraining.

cs.SE↗
arXiv2026

Thermodynamic Crossovers in Supercritical Matter: One, No One, and One Hundred Thousand

Beyond a critical point, two thermodynamic phases become continuously connected, without being separated by any thermodynamic singularity. The paradigmatic example is the liquid-gas transition, where several criteria have been proposed to distinguish liquid-like and gas-like regimes within the supercritical region. Recently, a pair of thermodynamic crossover lines, $L^\pm$, was introduced to address a proposed incompatibility of previous definitions with the universal critical behavior of the Ising model. These lines divide the supercritical region into three regimes, meet only at the critical point, and obey the expected critical scaling laws. Here, we demonstrate that these requirements do not uniquely determine the crossover boundaries. Instead, infinitely many pairs of lines satisfy the same conditions, with different choices of thermodynamic paths selecting different crossover boundaries. We further show that the Lee-Yang formalism, despite providing an intrinsic characterization of critical singularities, does not resolve this ambiguity in defining real-axis crossover boundaries. Restricting to regular probing paths that respect the $Z_2$ symmetry of the critical point recovers the Widom and $L^\pm$ prescriptions asymptotically. This restriction, however, is a choice of protocol rather than a property of the system, and even then two inequivalent prescriptions survive. Our results demonstrate that universal critical behavior alone does not select privileged supercritical thermodynamic crossover boundaries. While such lines may identify physically meaningful changes in thermodynamic, structural, or transport properties, their precise locations and the widths of the regions they delimit remain dependent on the operational criteria used to define them.

cond-mat.stat-mech↗
arXiv2026

Spectral Smoothing and Spiking Perturbations Reconfigure Capability Profiles and Selective Task Performance in Large Language Models

Singular-value spectra encode important structure in transformer feed-forward projections, yet their causal role in downstream behavior remains underexplored. We study post-hoc spectral smoothing and spiking interventions on the up_proj and down_proj matrices. Positive perturbation stretches deviations from the geometric mean, making large singular values larger and small ones smaller (spiking), and vice versa. The intervention preserves singular vectors and the geometric mean of the singular values. Across five open Llama and Qwen instruction models, we evaluate the perturbed systems on capability-labeled subtasks, standard and chain-of-thought reasoning benchmarks, judge-scored truthfulness, hallucination, and safety tasks, and TRAIT personality probes. To interpret score shifts, we fit responses in a three-dimensional capability basis consisting of factual knowledge, language understanding, and deductive reasoning. The induced changes suggest structured rather than purely random sensitivity: large positive perturbations more reliably suppress language and deductive responses, moderate negative perturbations can sometimes improve factual or task-local performance on the evaluated grids, and sensitivity varies strongly by model family, benchmark module, and underlying layerwise spectral inequality. These results position spectral perturbation as a lightweight intervention for probing how capabilities are organized and traded off in large language models.

cs.CL↗
arXiv2026

Spark-Audio 1.0 Technical Report

This paper introduces Spark-Audio 1.0, a speech foundation model trained on domestically produced computing infrastructure. The model accepts text and speech inputs and generates text outputs. It supports multilingual (99 languages) and multi-dialect (202 dialects) speech recognition, speech translation, audio perception and understanding, and both text- and speech-based dialogue, marking a shift from merely ``hearing clearly'' to genuinely ``understanding'' speech. Architecturally, Spark-Audio 1.0 couples a 0.65B-parameter dense audio encoder with a 30B-A3B Mixture-of-Experts (MoE) large language model (LLM). The audio encoder is pretrained with representation-enhanced BEST-RQ and then refined through multi-task supervised training. The audio--LLM is trained through staged pretraining, followed by post-training that combines supervised fine-tuning with on-policy distillation. In total, the model is trained on 13 million hours of audio data together with large-scale text data. We evaluate Spark-Audio 1.0 on mainstream audio and text benchmarks, as well as on in-house test sets drawn from real-world application scenarios. Despite using substantially less audio training data than Qwen3.5-Omni-Flash, Spark-Audio 1.0 achieves comparable performance on most audio and text tasks, outperforms it in multilingual speech recognition, spoken dialogue, and real-world application scenarios, and approaches the performance of larger closed-source models. Nevertheless, there remains room for improvement in instruction following and complex audio reasoning. Finally, the successful training of Spark-Audio 1.0 on domestic hardware, including Huawei Ascend 910B and Hygon BW1000 chips, demonstrates the feasibility of a domestic technical pathway for building speech foundation models.

eess.AS↗
arXiv2026

CoASFL: Coupled Consistency Optimization for Asynchronous Split Federated Learning

Split Federated Learning (SFL) has emerged as a promising paradigm for training large-scale deep models on resource-constrained edge devices. Although asynchronous SFL alleviates the straggler bottleneck caused by system heterogeneity, it inevitably introduces two coupled dimensions of inconsistency: horizontal inconsistency across clients and vertical inconsistency between clients and the server. These inconsistencies, further exacerbated by statistical heterogeneity, result in biased activation representations, model divergence, and severe client drift. To address these challenges, we propose CoASFL, an inconsistency-aware asynchronous SFL framework that incorporates a dynamic activation merging mechanism and a consistency-aware model aggregation mechanism. Specifically, CoASFL constructs more consistent activation sequences from heterogeneous clients to alleviate horizontal inconsistency, and dynamically adjusts aggregation weights according to client participation frequencies and progress gaps to mitigate vertical inconsistency. By jointly coordinating activation-level and model-level consistency, CoASFL effectively suppresses inconsistency propagation during asynchronous training. Extensive experiments on an 80-device heterogeneous edge testbed demonstrate that CoASFL consistently outperforms state-of-the-art SFL approaches, accelerating training by 1.61$\times$ to 3.74$\times$ and improving final model accuracy by up to 3.26\% under heterogeneous scenarios.

cs.LG↗
arXiv2026

ID Balancing: Stable Training of Extremely Sparse MoE via PID-Based Load Control

Scaling Large Language Models (LLMs) via Mixture-of-Experts (MoE) enables massive parameter growth with nearly constant per-token computation. However, further scaling the parameter count requires increasingly sparse routing, where expert load imbalance becomes more severe. This imbalance reduces parameter utilization and training efficiency, and can undermine training stability, becoming a bottleneck to reliable scaling. In this work, we unify two representative auxiliary-loss-free methods as incomplete Proportional-Integral-Derivative (PID) controllers: DeepSeek's loss-free method acts as a fixed-step integral controller, while Kimi K3's Quantile Balancing functions as a generalized proportional controller. Building on this control perspective, we propose ID Balancing, an Integral-Derivative controller. It scales its integral term with load error and activates its derivative term only when imbalance worsens, enabling stronger corrections for large or worsening errors and smaller updates near balance. Evaluated across Top-$10$, Top-$5$, and Top-$3$ routing over $768$ experts, ID Balancing reduces worst-case backbone MaxVio and training-average backbone MinVio by over $50$% and $12$%, respectively, relative to the best baselines in the Top-$3$ setting. When the total parameter count increases from $18.9$B to $69.9$B (Top-$10$-of-$768$), ID Balancing's worst-case backbone MaxVio remains nearly unchanged and is approximately $89.6$% lower than that of the auxiliary-loss baseline. ID Balancing also maintains competitive language-modeling and downstream performance. The advantages of ID Balancing grow as sparsity increases, making it a promising solution for scaling larger, sparser MoE models.

cs.LG↗
arXiv2026

Efficient CPU-GPU collaborative MoE inference via Adaptive Expert Assignment

Mixture-of-Experts (MoE) models reduce inference computation through sparse expert activation, but their large expert memory footprint often exceeds GPU capacity, making expert offloading necessary for memory-constrained inference. However, existing systems typically handle GPU-missed experts by either transferring them to the GPU, incurring substantial PCIe transfer latency, or executing them on the CPU, making the slower CPU computation the bottleneck. To address these limitations, we present \method, a CPU-GPU collaborative MoE inference system that adaptively assigns these missed experts to the CPU or GPU for efficient execution. By exploiting differences in expert token-loads, \method selectively transfers high-load experts to the GPU while executing low-load experts on the CPU. To minimize inference latency, \method adaptively adjusts expert assignments to balance the CPU computation path against the GPU transfer-and-computation path, allowing both execution paths to complete as early as possible. Extensive experiments on three representative MoE LLMs show that \method improves prefill speed by up to 51.9\% and decode throughput by up to 39.5\% over the strongest baseline across the evaluated batch sizes, with consistent gains under different GPU-memory budgets.

cs.AI↗
arXiv2026

Deploying RDMA for Distributed Inference and Training At Scale

RDMA can reduce communication overhead in distributed machine learning systems, but realizing these benefits in production requires more than replacing TCP. Variable-size messages, concurrent connections, bursty incasts, and service failures introduce challenges that hardware offload alone cannot address. This paper reports our experience deploying ByteExpress, a software communication layer over hardware RDMA, in ByteDance's parameter-server-based inference and training systems. We present three lessons from the deployment: efficient RDMA requires buffer management across application and transport layers; scalable communication requires coordinated management of connections, receive capacity, and outstanding data; and service availability requires recovery above the transport layer. ByteExpress combines an RDMA-aware allocator, zero-copy transfers, QP multiplexing, receive-credit flow control, a software send window, and RPC-assisted recovery. In production inference, ByteExpress reduces mean and P99 model parameter fetch latency by 35.7% and 43.7%, respectively, and total CPU usage by 10.69% compared with kernel TCP. In training, ByteExpress reduces P99 model parameter push latency by up to 74.0% compared with TCP, with no PFC pauses observed at runtime.

cs.LG↗
arXiv2026

Deep Jump Gaussian Processes for Surrogate Modeling of High-Dimensional Piecewise Continuous Functions

We introduce Deep Jump Gaussian Processes (DJGP), a novel method for surrogate modeling of a piecewise continuous function on a high-dimensional domain. DJGP addresses the limitations of conventional Jump Gaussian Processes (JGP) in high-dimensional input spaces by integrating region-specific, locally linear projections with JGP modeling. These projections employ region-dependent matrices to capture local low-dimensional subspace structures, making them well suited to the inherently localized modeling behavior of JGPs, a variant of local Gaussian processes. To control model complexity, we place a Gaussian Process prior on the projection matrices, allowing them to evolve smoothly across the input space. The projected inputs are then modeled with a JGP to capture piecewise continuous relationships with the response. This yields a distinctive two-layer deep learning of GP/JGP. We further develop a scalable variational inference algorithm to jointly learn the projection matrices and JGP hyperparameters. Rigorous theoretical analysis and extensive empirical studies are provided to justify the proposed approach. In particular, we derive an oracle error bound for DJGP and decompose it into four distinct sources of error, which are then linked to practical implications. Experiments on synthetic and benchmark datasets demonstrate that DJGP achieves superior predictive accuracy and more reliable uncertainty quantification compared with existing methods.

cs.LG↗
arXiv2026

Uniform Information Density-Based Preference Optimization

To enable large language models to generate outputs that align with human preferences, optimization algorithms based on the comparison between chosen and rejected output pairs have been developed and been proven effective. Some recent endeavors have utilized certain linguistically-derived assumptions about the preference pairs to guide the design of more precise optimization directions. However, human preferences are complex and hard to measure by simplistic criteria, and it is important to incorporate broader and more sophisticated insights about preferred vs. not-preferred language. In this study, we adopt a well-established psycholinguistic theory, Uniform Information Density, to develop a novel preference optimization algorithm, UIDPO. The basic assumption is that humans prefer to produce language in which information is uniformly distributed, because uniformity induces less cognitive processing effort. To implement this idea into the algorithm, we derive a compound metric for information uniformity by combining the global standard deviation and the local second-order difference, and integrate it to the objective of preference optimization. We compare UIDPO with SimPO, a previous state-of-the-art method, obtaining comparable and sometimes better performance on AlpacaEval2, ArenaHard, and MT-Bench. Additional linguistic analysis on complement clauses also shows that UIDPO can better reflect the tendencies of keeping information uniform in natural language than baselines. Our results show that the subtle human preference over the information rate in language is a helpful source for achieving better human alignment for LLMs. The code repository is published here: https://github.com/CLCS-SUSTech/UIDPO.

cs.CL↗
arXiv2026

Modeling the Arrival Time of Information-Bearing Events in Natural Language

Understanding how information is distributed over time is fundamental to modeling natural language. While existing empirical studies primarily rely on post-hoc analysis of information-theoretic metrics, such as surprisal, little is known about the generation processes under the surface. We propose a novel approach for addressing this gap using the tool of neural temporal point processes (TPPs) by modeling the key information-bearing events in text sequences with learnable event intensity functions. Our model takes the history of event intervals and the contextual semantic representations from LLMs as input, and predicts the future arrival times for two types of events: new-word and high-surprisal-token. The experimental results reveal an interesting distinction in event intensities: LLMs tend to produce new words at significantly higher rates than humans, while high-surprisal tokens do not exhibit the difference. A decaying pattern of intensity over time is found in new-word events for both humans and LLMs, but not in high-surprisal-token events. Our findings revisit and extend previous studies that use simpler methods; provide supplementary evidence for theories about constrained context representation inherent in human language production; and find a new angle to view human and LLM generation differences.

cs.CL↗
arXiv2026

Generalized Rank Weight and Extended Generalized Poset Weight Defined For Codes Over Rings: A Galois Connection Approach

In this paper, we study generalized rank weights (GRWs) and extended generalized poset weight (EGPWs) of codes over rings via a Galois connection approach. First, we show that various coding-theoretic properties related to generalized weights, including security drops of a code employed in wire-tap channel of type II, connections between generalized weights of a Gabidulin code and its associated Delsarte code, (generalized) Singleton bound, MDS discrepancy of a code, characterizations of MDS, near MDS, $i$-MDS, MRD, near MRD, $i$-MRD, (dually) quasi-MRD codes as well as evasive property of subspaces, can be reformulated in terms of Galois connections. Next, we study GRWs and rank profiles defined for modules over principal ideal rings, especially those over chain rings. Generalizing GRWs defined for vector spaces over fields, we establish a singleton bound and a Wei-type duality theorem, characterize MRD, near MRD and dually quasi-MRD codes and determine their GRWs; moreover, we characterize $i$-MRD codes and establish a scattered bound for $(h,h)$-evasive codes over chain rings, generalizing counterpart result established for vector space over finite fields. Finally, we propose and study EGPWs and extended poset profiles defined for modules with a composition series, which in fact form a Galois connection. Generalizing EGPWs defined for modules over finite Galois rings, we establish a Wei-type duality theorem for modules over arbitrary quasi-Frobenius rings, which unifies the two Wei-type duality theorems derived in both \cite{32} and \cite{33}.

cs.IT↗
arXiv2026

$r$-Minimal Poset Codes

In this paper, we propose and study $r$-minimal codes with respect to $\mathbf{P}$-support, where $\mathbf{P}=(Ω,\preccurlyeq_{\mathbf{P}})$ is a poset defined on the coordinate set of the ambient space $\mathbf{H}$. $r$-Minimal $\mathbf{P}$-codes are natural extensions of Hamming metric minimal codes that have been extensively studied in the literature. We characterize $r$-minimal $\mathbf{P}$-codes in terms of the notion so called cutting $r$-blocking maps, which generalizes the well-known equivalence between minimal Hamming metric codes and cutting blocking sets. We also give a necessary and sufficient condition for $r$-minimality in terms of $(\mathbf{P},ω)$-weight defined on $\mathbf{H}$, where $ω:Ω\longrightarrow\mathbb{R}^{+}$ is an arbitrary weight function. This leads to a generalization of the well-known Ashikhmin-Barg criterion for Hamming metric minimal codes. We then prove two existence results for $r$-minimal $\mathbf{P}$-codes, both for general $\mathbf{P}$ and for the special case that $\mathbf{P}$ is a disjoint union of chains. When $\mathbf{P}$ is hierarchical, we characterize $r$-minimal $\mathbf{P}$-codes in terms of $r$-minimal Hamming metric codes. Finally, we characterize cutting $r$-blocking sets induced by hierarchical posets with two levels, which further enables us to answer a question raised in Hyun, Kim, Wu and Yue \cite{28}.

cs.IT↗
arXiv2026

Programmable nonlinearity within nanophotonic waveguides

Nonlinear nanophotonics enables the engineering of optical functions using light-matter interactions in nanoscale structures with carefully designed geometries. Phase matching in nonlinear optics is typically essential for realizing high efficiency for desired processes and engineering of sophisticated functions, but it can be challenging to reliably achieve in nanophotonic waveguides. Conventionally, phase matching in nanophotonics is achieved through nanofabrication, e.g., by poling a device's nonlinear material using electrodes with fixed geometry. This approach has two drawbacks: the optical functions a device can perform are fixed at the time of fabrication, and the device's performance can be degraded by fabrication imperfections. Here, we circumvent these limitations by engineering programmable $χ^{(2)}$ holograms inside nanophotonic waveguides, enabling reconfigurable quasi-phase matching that can be updated after device fabrication in approximately one second. By projecting structured optical illumination onto photoconductive electrodes, we generated spatial patterns of electric-field-induced $χ^{(2)}$ to exert spectral, modal, and polarization control over three-wave mixing processes. This programmability also allowed us to employ closed-loop in situ optimization to compensate for phase mismatch in complex waveguide structures, such as width-modulated waveguides and spiral waveguides exceeding 10 cm in length. Furthermore, we show that programmable nonlinearity enables powerful in situ diagnostics, including direct dispersion measurements of various transverse modes and on-chip wavefront phase tomography. Our work opens up the possibility of building cascaded nonlinear nanophotonic systems where yield issues due to fabrication variations can be sidestepped while retaining the benefits of etched waveguides such as confinement and dispersion engineering.

physics.optics↗