arXiv subjects

Wen Wu

Publications and source records attributed to Wen Wu.

At least 55 records · Page 3Linked to original sources

arXiv2026

Drag penalty during relaminarization and Kelvin-Helmholtz-promoted retransition in an accelerating turbulent boundary layer over initially drag-reducing riblets

Direct numerical simulations of an accelerating turbulent boundary layer (TBL) over a smooth wall and a wall fully covered with streamwise-aligned riblets are performed to investigate drag modulation and its underlying mechanisms. The riblet-scale flow is resolved using an immersed boundary method. Starting from a zero-pressure-gradient (ZPG) TBL at Re=6800, the flow undergoes a threefold freestream acceleration over seventy-five boundary-layer thicknesses, matching the development reported by Warnack and Fernholz (1998), and consequently experiences relaminarization followed by retransition farther downstream. The riblets, defined by a sinusoidal spanwise profile with initial s+=15.2 and lg+=10.5, correspond to near-optimal drag-reducing size in ZPG flows. However, even modest acceleration renders them drag-increasing, showing that the conventional ZPG interpretation based on total-drag viscous scaling does not apply directly in this non-equilibrium flow. During relaminarization, the drag penalty arises primarily from geometry-determined concentration of viscous shear near the riblet crest, with negligible direct Reynolds- and dispersive-stress contributions prior to retransition. Despite the drag increase, the overlying TBL remains statistically similar to the smooth-wall case when scaled with the total shear stress at the groove opening, demonstrating that this shear sets the relevant scaling for the TBL, while the additional drag generated within the grooves remains largely decoupled from the outer-layer turbulence dynamics. This partial decoupling persists until the onset of retransition, when spanwise Kelvin-Helmholtz rollers develop near the riblet crest and promote earlier, stronger retransition through their interaction with the residual near-wall streaks. These findings provide a revised physical picture of riblet performance in non-equilibrium turbulent flows.

physics.flu-dyn↗

arXiv2026

Fast AI Model Partition for Split Learning over Edge Networks

Split learning (SL) is a distributed learning paradigm that can enable computation-intensive artificial intelligence (AI) applications by partitioning AI models between mobile devices and edge servers. %fully utilizing distributed computing resources for computation-intensive mobile intelligence applications. However, the model partitioning problem in SL becomes challenging due to the diverse and complex architectures of AI models. In this paper, we formulate an optimal model partitioning problem to minimize training delay in SL. To solve the problem, we represent an arbitrary AI model as a directed acyclic graph (DAG), where the model's layers and inter-layer connections are mapped to vertices and edges, and training delays are captured as edge weights. Then, we propose a general model partitioning algorithm by transforming the problem into a minimum \textit{s-t} cut problem on the DAG. Theoretical analysis shows that the two problems are equivalent, such that the optimal model partition can be obtained via a maximum-flow method. Furthermore, taking AI models with block structures into consideration, we design a low-complexity block-wise model partitioning algorithm to determine the optimal model partition. Specifically, the algorithm simplifies the DAG by abstracting each block (i.e., a repeating component comprising multiple layers in an AI model) into a single vertex. Extensive experimental results on a hardware testbed equipped with NVIDIA Jetson devices demonstrate that the proposed solution can reduce algorithm running time by up to 13.0$\times$ and training delay by up to 38.95\%, compared to state-of-the-art baselines.

cs.LG↗

arXiv2026

Communication-Efficient Collaborative LLM Inference over LEO Satellite Networks

Low Earth orbit (LEO) satellites play an essential role in intelligent Earth observation by leveraging artificial intelligence models. However, limited onboard memory and excessive inference delay prevent the practical deployment of large language models (LLMs) on a single satellite. In this paper, we propose a communication-efficient collaborative LLM inference scheme for LEO satellite networks. Specifically, the entire LLM is split into multiple sub-models, with each deployed on a satellite, thereby enabling collaborative LLM inference via exchanging intermediate activations between satellites. The proposed scheme also leverages the pipeline parallelism mechanism that overlaps sub-model inference with intermediate activation transmission, thereby reducing LLM inference delay. An adaptive activation compression scheme is designed to mitigate cumulative errors from multi-stage model splitting while preserving inference accuracy. Furthermore, we formulate the LLM inference delay minimization problem by jointly optimizing model splitting and compression ratios under onboard memory and inference accuracy constraints. The problem is transformed into a shortest-path search problem over a directed acyclic graph that edge weights explicitly quantify the inference delay induced by model splitting and compression strategies, which is solved via a modified A Star-based search algorithm. Extensive simulation results indicate that the proposed solution can reduce inference delay by up to 42% and communication overhead by up to 71% compared to state-of-the-art benchmarks, while maintaining the inference accuracy loss of less than 1%.

cs.DC↗

arXiv2026

Joint Shape-Position Optimization Enhanced 2D DOA Estimation in Movable Antenna Systems

Movable Antenna (MA) technology is emerging as a promising advancement with the potential to significantly enhance the performance of future wireless communication and sensing systems. In this paper, we address two-dimensional (2D) direction of arrival (DOA) estimation via joint shape-position optimization. Specifically, we formulate an optimization problem aimed at minimizing the Cramér-Rao Bound (CRB) based on a 2D DOA estimation model for MA systems. To tackle the highly non-convex nature of this CRB minimization, we investigate the spatial utilization of the movable region (MR) under minimum antenna spacing constraints. By demonstrating that an equilateral triangle yields the minimum overlap area, we strategically design an equilateral triangular MR. This specific geometric configuration enables the exploitation of structural symmetry to simplify the geometric constraints, which effectively reduces the complexity of solving the optimization problem. Subsequently, we derive the optimal MA positions by selecting the candidate locations farthest from the centroid of MR. The results demonstrate that the proposed joint shape-position optimization substantially enhances 2D DOA estimation performance.

eess.SP↗

arXiv2026

Intern-S1-Pro: Scientific Multimodal Foundation Model at Trillion Scale

We introduce Intern-S1-Pro, the first one-trillion-parameter scientific multimodal foundation model. Scaling to this unprecedented size, the model delivers a comprehensive enhancement across both general and scientific domains. Beyond stronger reasoning and image-text understanding capabilities, its intelligence is augmented with advanced agent capabilities. Simultaneously, its scientific expertise has been vastly expanded to master over 100 specialized tasks across critical science fields, including chemistry, materials, life sciences, and earth sciences. Achieving this massive scale is made possible by the robust infrastructure support of XTuner and LMDeploy, which facilitates highly efficient Reinforcement Learning (RL) training at the 1-trillion parameter level while ensuring strict precision consistency between training and inference. By seamlessly integrating these advancements, Intern-S1-Pro further fortifies the fusion of general and specialized intelligence, working as a Specializable Generalist, demonstrating its position in the top tier of open-source models for general capabilities, while outperforming proprietary models in the depth of specialized scientific tasks.

cs.LG↗

arXiv2026

STEP: Scientific Time-Series Encoder Pretraining via Cross-Domain Distillation

Scientific time series are central to scientific AI but are typically sparse, highly heterogeneous, and limited in scale, making unified representation learning particularly challenging. Meanwhile, foundation models pretrained on relevant time series domains such as audio, general time series, and brain signals contain rich knowledge, but their applicability to scientific signals remains underexplored. In this paper, we investigate the transferability and complementarity of foundation models from relevant time series domains, and study how to effectively leverage them to build a unified encoder for scientific time series. We first systematically evaluate relevant foundation models, showing the effectiveness of knowledge transfer to scientific tasks and their complementary strengths. Based on this observation, we propose STEP, a Scientific Time Series Encoder Pretraining framework via cross domain distillation. STEP introduces adaptive patching to handle extreme-length sequences and a statistics compensation scheme to accommodate diverse numerical scales. It further leverages cross-domain distillation to integrate knowledge from multiple foundation models into a unified encoder. By combining complementary representations across different domains, STEP learns general-purpose and transferable features tailored for scientific signals. Experiments on seven scientific time series tasks demonstrate that STEP provides both an effective structure and an effective pretraining paradigm, taking a STEP toward scientific time series representation learning.

cs.LG↗

arXiv2026

CAST-TTS: A Simple Cross-Attention Framework for Unified Timbre Control in TTS

Current Text-to-Speech (TTS) systems typically use separate models for speech-prompted and text-prompted timbre control. While unifying both control signals into a single model is desirable, the challenge of cross-modal alignment often results in overly complex architectures and training objective. To address this challenge, we propose CAST-TTS, a simple yet effective framework for unified timbre control. Features are extracted from speech prompts and text prompts using pre-trained encoders. The multi-stage training strategy efficiently aligns the speech and projected text representations within a shared embedding space. A single cross-attention mechanism then allows the model to use either of these representations to control the timbre. Extensive experiments validate that the unified cross-attention mechanism is critical for achieving high-quality synthesis. CAST-TTS achieves performance comparable to specialized single-input models while operating within a unified architecture. The demo page can be accessed at https://HiRookie9.github.io/CAST-TTS-Page.

cs.SD↗

arXiv2026

Improving boundary-layer separation prediction by an IDDES turbulence model using a pressure-gradient sensor

This work extends a pressure-gradient sensor for boundary-layer separation originally developed for the $k-ω$ shear-stress transport Reynolds-averaged Navier-Stokes (RANS) model (Griffin et al., 2025, J. Turb.) to the Improved Delayed Detached Eddy Simulation (IDDES) turbulence model of Gritskevich et al. (2012, Flow Turbul. Combust.). The pressure-gradient sensor identifies local regions of strong adverse pressure-gradient where the eddy-viscosity is reduced, as in the original RANS model. Additionally, to promote separation in the IDDES model, the elevation term in the IDDES length scale, designed to augment the RANS-mode Reynolds stress in attached flow regions, is turned off where the pressure-gradient sensor is active. The model is applied on various airfoils representative of both wind energy and aerospace applications, and is used in fully turbulent and transitional IDDES model variants. The proposed model improves the prediction of stall onset and post-stall regimes relative to the baseline IDDES model without significant degradation of attached-flow regimes relative to state-of-the-art RANS models or deep-stall regimes relative to state-of-the-art IDDES models. Significant overall improvements are observed in predictions of lift and drag polars across 90 degrees of angle of attack, yielding a unified model able to predict various (two- and three-dimensional) flow regimes. Shortcomings are identified to be related to the underlying RANS pressure-gradient sensor rather than the extension of the sensor to the IDDES model, which is the focus of this work.

physics.flu-dyn↗

arXiv2026

SciTS: Scientific Time Series Understanding and Generation with LLMs

The scientific reasoning ability of large language models (LLMs) has recently attracted significant attention. Time series, as a fundamental modality in scientific data, presents unique challenges that are often overlooked in current multimodal LLMs, which either encode numerical sequences as text or convert them into images. Such approaches may be insufficient for comprehensive scientific time series understanding and generation. Existing unified time series models typically specialise in either forecasting or analysis, and their effectiveness on non-periodic, heterogeneous scientific signals remains unclear. To address these gaps, we introduce SciTS, a benchmark spanning 12 scientific domains and 43 tasks, with over 50k+ instances, both univariate and multivariate signals ranging from $10^0$ to $10^7$ in length and up to 10~MHz in frequency. We benchmark 17 models, including text-only LLMs, multimodal LLMs, and unified time series models, and find that general-purpose LLMs exhibit stronger generalisability than specialised time series models, while representing time series as text or images limits their performance due to excessively long sequences and loss of numerical precision, respectively. We then introduce TimeOmni, a framework that equips LLMs with the ability to understand and generate time series while remaining compatible with general-purpose LLM training. This work fills a gap in both dedicated benchmarks and modelling frameworks for scientific time series, paving the way for LLMs to understand and generate complex temporal scientific data.

cs.LG↗

arXiv2026

One Brain, Omni Modalities: Towards Unified Non-Invasive Brain Decoding with Large Language Models

Deciphering brain function through non-invasive recordings requires synthesizing complementary high-frequency electromagnetic (EEG/MEG) and low-frequency metabolic (fMRI) signals. However, despite their shared neural origins, extreme discrepancies have traditionally confined these modalities to isolated analysis pipelines, hindering a holistic interpretation of brain activity. To bridge this fragmentation, we introduce \textbf{NOBEL}, a \textbf{n}euro-\textbf{o}mni-modal \textbf{b}rain-\textbf{e}ncoding \textbf{l}arge language model (LLM) that unifies these heterogeneous signals within the LLM's semantic embedding space. Our architecture integrates a unified encoder for EEG and MEG with a novel dual-path strategy for fMRI, aligning non-invasive brain signals and external sensory stimuli into a shared token space, then leverages an LLM as a universal backbone. Extensive evaluations demonstrate that NOBEL serves as a robust generalist across standard single-modal tasks. We also show that the synergistic fusion of electromagnetic and metabolic signals yields higher decoding accuracy than unimodal baselines, validating the complementary nature of multiple neural modalities. Furthermore, NOBEL exhibits strong capabilities in stimulus-aware decoding, effectively interpreting visual semantics from multi-subject fMRI data on the NSD and HAD datasets while uniquely leveraging direct stimulus inputs to verify causal links between sensory signals and neural responses. NOBEL thus takes a step towards unifying non-invasive brain decoding, demonstrating the promising potential of omni-modal brain understanding.

q-bio.NC↗

arXiv2026

HoliAntiSpoof: Audio LLM for Holistic Speech Anti-Spoofing

Recent advances in speech synthesis and editing have made speech spoofing increasingly challenging. However, most existing methods treat spoofing as binary classification, overlooking that diverse spoofing techniques manipulate multiple, coupled speech attributes and their semantic effects. In this paper, we introduce HoliAntiSpoof, the first audio large language model (ALLM) framework for holistic speech anti-spoofing analysis. HoliAntiSpoof reformulates spoofing analysis as a unified text generation task, enabling joint reasoning over spoofing methods, affected speech attributes, and their semantic impacts. To support semantic-level analysis, we introduce DailyTalkEdit, a new anti-spoofing benchmark that simulates realistic conversational manipulations and provides annotations of semantic influence. Extensive experiments demonstrate that HoliAntiSpoof outperforms conventional baselines across multiple settings, while preliminary results show that in-context learning further improves out-of-domain generalization. These findings indicate that ALLMs not only enhance speech spoofing detection performance but also enable interpretable analysis of spoofing behaviors and their semantic effects, pointing towards more trustworthy and explainable speech security. Data and code are publicly available.

cs.SD↗

arXiv2026

Speaker Anonymisation for Speech-based Suicide Risk Detection

Adolescent suicide is a critical global health issue, and speech provides a cost-effective modality for automatic suicide risk detection. Given the vulnerable population, protecting speaker identity is particularly important, as speech itself can reveal personally identifiable information if the data is leaked or maliciously exploited. This work presents the first systematic study of speaker anonymisation for speech-based suicide risk detection. A broad range of anonymisation methods are investigated, including techniques based on traditional signal processing, neural voice conversion, and speech synthesis. A comprehensive evaluation framework is built to assess the trade-off between protecting speaker identity and preserving information essential for suicide risk detection. Results show that combining anonymisation methods that retain complementary information yields detection performance comparable to that of original speech, while achieving protection of speaker identity for vulnerable populations.

eess.AS↗

arXiv2026

MMEDIT: A Unified Framework for Multi-Type Audio Editing via Audio Language Model

Text-guided audio editing aims to modify specific acoustic events while strictly preserving non-target content. Despite recent progress, existing approaches remain fundamentally limited. Training-free methods often suffer from signal degradation caused by diffusion inversion, while training-based methods, although achieving higher generation quality, are severely constrained by the scarcity of high-quality paired data and task formulations that cover only a narrow subset of editing operations. In addition, standard architectures typically decouple text and audio processing, limiting the ability to align instructions with specific acoustic contexts. To address these challenges, we propose MMEdit, an audio-language-model-driven framework for unified audio editing. We systematically extend task definitions to cover a comprehensive range of editing operations, including addition, replacement, removal, reordering, and attribute modification. Furthermore, we design a scalable data synthesis pipeline to construct large-scale paired datasets with fine-grained event-level annotations. To capture complex editing semantics, we integrate a Qwen2-Audio encoder with an MMDiT-based generator, enabling precise cross-modal alignment and localized editing. Experimental results demonstrate that our method achieves superior editing localization accuracy, robust instruction following, and high fidelity in non-edited regions.

cs.SD↗

arXiv2026

HIPPO: Accelerating Video Large Language Models Inference via Holistic-aware Parallel Speculative Decoding

Speculative decoding (SD) has emerged as a promising approach to accelerate LLM inference without sacrificing output quality. Existing SD methods tailored for video-LLMs primarily focus on pruning redundant visual tokens to mitigate the computational burden of massive visual inputs. However, existing methods do not achieve inference acceleration comparable to text-only LLMs. We observe from extensive experiments that this phenomenon mainly stems from two limitations: (i) their pruning strategies inadequately preserve visual semantic tokens, degrading draft quality and acceptance rates; (ii) even with aggressive pruning (e.g., 90% visual tokens removed), the draft model's remaining inference cost limits overall speedup. To address these limitations, we propose HIPPO, a general holistic-aware parallel speculative decoding framework. Specifically, HIPPO proposes (i) a semantic-aware token preservation method, which fuses global attention scores with local visual semantics to retain semantic information at high pruning ratios; (ii) a video parallel SD algorithm that decouples and overlaps draft generation and target verification phases. Experiments on four video-LLMs across six benchmarks demonstrate HIPPO's effectiveness, yielding up to 3.51x speedup compared to vanilla auto-regressive decoding.

cs.CV↗

arXiv2025

Application of riblets to separating turbulent boundary layers

We conduct direct numerical simulations of separating turbulent boundary layers (TBLs) over triangular riblets with tip angles $90^o$ (T9) and $60^o$ (T6). Our setup follows the separating TBL study of Wu et al.\ ({\it J. Fluid Mech.}, vol.\ 883, 2020, p.\ A45). An equilibrium zero pressure-gradient (ZPG) TBL is generated at a reference location, followed by imposition of a Gaussian suction profile to create a separation bubble. The ZPG TBLs over the riblets and the benchmark smooth case have matched momentum thickness Reynolds number $Re_{θ_0} = 583$ (friction Reynolds number 224). We employ a well-validated spectral-element solver, and leverage its unstructured-grid nature to generate an optimal grid, based on the size of turbulent scales across the TBL. At the reference location, the T9 and T6 riblets respectively increase and reduce drag, with viscous-scaled spacings $52$ and $13$. We discover that for both riblet cases, the mean separation location occurs at a distance of $140θ_0$ downstream of the reference location, $18\%$ shorter than the mean separation distance for the smooth case ($170θ_0$). This outcome is related to the progressive enhancement of the Kelvin-Helmholtz (KH) rollers over the riblets, owing to the continuous rise in the adverse pressure-gradient. The KH rollers penetrate into the turbulent separation bubble, with significantly larger size and coherence compared to their counterparts upstream of the mean separation location.

physics.app-ph↗

arXiv2025

Hausdorff dimension of Graphs of Limit Functions Generated by Quasi-Linear Functions

The limit functions generated by quasi-linear functions or sequences (including the sum of the Rudin-Shapiro sequence as an example) are continuous but almost everywhere non-differentiable functions. Their graphs are fractal curves. In 2017 and 2020, Chen, Lü, Wen and the first author studied the box dimension of the graphs of the limit functions. In this paper, we focus on the Hausdorff dimension of the graphs of such limit functions. We first prove that the Hausdorff dimension of the graph of the limit function generated by the abelian complexity of the Rudin-Shapiro sequence is $\frac{3}{2}$. Then we extend the result to the graphs of limit functions generated by quasi-linear functions.

math.MG↗

arXiv2025

BrainOmni: A Brain Foundation Model for Unified EEG and MEG Signals

Electroencephalography (EEG) and magnetoencephalography (MEG) measure neural activity non-invasively by capturing electromagnetic fields generated by dendritic currents. Although rooted in the same biophysics, EEG and MEG exhibit distinct signal patterns, further complicated by variations in sensor configurations across modalities and recording devices. Existing approaches typically rely on separate, modality- and dataset-specific models, which limits the performance and cross-domain scalability. This paper proposes BrainOmni, the first brain foundation model that generalises across heterogeneous EEG and MEG recordings. To unify diverse data sources, we introduce BrainTokenizer,the first tokenizer that quantises spatiotemporal brain activity into discrete representations. Central to BrainTokenizer is a novel Sensor Encoder that encodes sensor properties such as spatial layout, orientation, and type, enabling compatibility across devices and modalities. Building upon the discrete representations, BrainOmni learns unified semantic embeddings of brain signals by self-supervised pretraining. To the best of our knowledge, it is the first foundation model to support both EEG and MEG signals, as well as the first to incorporate large-scale MEG pretraining. A total of 1,997 hours of EEG and 656 hours of MEG data are curated and standardised from publicly available sources for pretraining. Experiments show that BrainOmni outperforms both existing foundation models and state-of-the-art task-specific models on a range of downstream tasks. It also demonstrates strong generalisation to unseen EEG and MEG devices. Further analysis reveals that joint EEG-MEG (EMEG) training yields consistent improvements across both modalities. Code and checkpoints are publicly available at https://github.com/OpenTSLab/BrainOmni.

eess.SP↗

arXiv2025

PicoAudio2: Temporal Controllable Text-to-Audio Generation with Natural Language Description

While recent work in controllable text-to-audio (TTA) generation has achieved fine-grained control through timestamp conditioning, its scope remains limited by audio quality and input format. These models often suffer from poor audio quality in real datasets due to sole reliance on synthetic data. Moreover, some models are constrained to a closed vocabulary of sound events, preventing them from controlling audio generation for open-ended, free-text queries. This paper introduces PicoAudio2, a framework that advances temporal-controllable TTA by mitigating these data and architectural limitations. Specifically, we use a grounding model to annotate event timestamps of real audio-text datasets to curate temporally-strong real data, in addition to simulation data from existing works. The model is trained on the combination of real and simulation data. Moreover, we propose an enhanced architecture that integrates the fine-grained information from a timestamp matrix with coarse-grained free-text input. Experiments show that PicoAudio2 exhibits superior performance in terms of temporal controllability and audio quality.

cs.SD↗