arXiv subjects

Rui Zhang

Publications and source records attributed to Rui Zhang.

At least 253 records · Page 14Linked to original sources

arXiv2026

Two-stage Multi-beam Training for Multiuser Millimeter-Wave Communications

In this letter, we study an efficient multi-beam training method for multiuser millimeter-wave communication systems. Unlike the conventional single-beam training method that relies on exhaustive search, multi-beam training design faces a key challenge in balancing the trade-off between beam training overhead and success beam-identification rate, exacerbated by severe inter-beam interference. To tackle this challenge, we propose a new two-stage multi-beam training method with two distinct multi-beam patterns to enable fast and accurate user angle identification. Specifically, in the first stage, the antenna array is divided into sparse subarrays to generate multiple beams (with high array gains), for identifying candidate user angles. In the second stage, the array is redivided into dense subarrays to generate flexibly steered wide beams, for which a cross-validation method is employed to effectively resolve the remaining angular ambiguity in the first stage. Last, numerical results demonstrate that the proposed method significantly improves the success beam-identification rate compared to existing multi-beam training methods, while retaining or even reducing the required beam training overhead.

eess.SP↗

arXiv2025

A Tutorial on MIMO-OFDM ISAC: From Far-Field to Near-Field

Integrated sensing and communication (ISAC) is one of the key usage scenarios for future sixth-generation (6G) mobile communication networks, where communication and sensing (C&S) services are simultaneously provided through shared wireless spectrum, signal processing modules, hardware, and network infrastructure. Such an integration is strengthened by the technology trends in 6G, such as denser network nodes, larger antenna arrays, wider bandwidths, higher frequency bands, and more efficient utilization of spectrum and hardware resources, which incentivize and empower enhanced sensing capabilities. As the dominant waveform used in contemporary communication systems, orthogonal frequency division multiplexing (OFDM) is still expected to be a very competitive technology for 6G, rendering it necessary to thoroughly investigate the potential and challenges of OFDM ISAC. Thus, this paper aims to provide a comprehensive tutorial overview of ISAC systems enabled by large-scale multi-input multi-output (MIMO) and OFDM technologies and to discuss their fundamental principles, advantages, and enabling signal processing methods. To this end, a unified MIMO-OFDM ISAC system model is first introduced, followed by four frameworks for estimating parameters across the spatial, delay, and Doppler domains, including parallel one-domain, sequential one-domain, joint two-domain, and joint three-domain parameter estimation. Next, sensing algorithms and performance analyses are presented in detail for far-field scenarios where uniform plane wave (UPW) propagation is valid, followed by their extensions to near-field scenarios where uniform spherical wave (USW) characteristics need to be considered. Finally, this paper points out open challenges and outlines promising avenues for future research on MIMO-OFDM ISAC.

eess.SP↗

arXiv2025

Frequency-switching Array Enhanced Physical-Layer Security in Terahertz Bands: A Movable Antenna Perspective

In this paper, we propose a new frequency-switching array (FSA) to enhance the physical-layer security (PLS) in the presence of multiple eavesdroppers (Eves), where the carrier frequency can be flexibly switched and small frequency offsets can be imposed on each antenna at the secrecy transmitter (Alice).First, we analytically show that by flexibly controlling the carrier frequency parameters, FSAs can effectively form uniform/non-uniform sparse arrays, hence resembling existing mechanically controlled movable antennas (MAs) via the control of inter-antenna spacing and providing additional degree-of-freedom in the beam manipulation.Although the proposed FSA suffers from additional path-gain attenuation in the received signals, it can overcome several hardware and signal processing issues incurred by MAs, such as limited positioning accuracy, extra hardware and energy cost.Then, a secrecy-rate maximization problem is formulated under the constraints on the frequency control.To shed useful insights, we first consider a secrecy-guaranteed problem with a null-steering constraint for which maximum ratio transmission beamformer is considered at Alice and the frequency offsets are set as uniform frequency increment.Interestingly, it is shown that the proposed FSA can flexibly realize null-steering over Eve in both the angular domain and range domain, thereby achieving improved PLS performance.Then, for the general case, we propose an efficient algorithm to solve the formulated non-convex optimization problem by using the block coordinate descent and projected gradient ascent techniques. Finally, numerical results demonstrate that the proposed FSA achieves superior secrecy rate performance over conventional fixed-position array, while it only suffers a slight secrecy rate loss than the existing mechanically controlled MA.

eess.SP↗

arXiv2025

StepFun-Formalizer: Unlocking the Autoformalization Potential of LLMs through Knowledge-Reasoning Fusion

Autoformalization aims to translate natural-language mathematical statements into a formal language. While LLMs have accelerated progress in this area, existing methods still suffer from low accuracy. We identify two key abilities for effective autoformalization: comprehensive mastery of formal-language domain knowledge, and reasoning capability of natural language problem understanding and informal-formal alignment. Without the former, a model cannot identify the correct formal objects; without the latter, it struggles to interpret real-world contexts and map them precisely into formal expressions. To address these gaps, we introduce ThinkingF, a data synthesis and training pipeline that improves both abilities. First, we construct two datasets: one by distilling and selecting large-scale examples rich in formal knowledge, and another by generating informal-to-formal reasoning trajectories guided by expert-designed templates. We then apply SFT and RLVR with these datasets to further fuse and refine the two abilities. The resulting 7B and 32B models exhibit both comprehensive formal knowledge and strong informal-to-formal reasoning. Notably, StepFun-Formalizer-32B achieves SOTA BEq@1 scores of 40.5% on FormalMATH-Lite and 26.7% on ProverBench, surpassing all prior general-purpose and specialized models.

cs.CL↗

arXiv2025

Unsupervised Single-Channel Audio Separation with Diffusion Source Priors

Single-channel audio separation aims to separate individual sources from a single-channel mixture. Most existing methods rely on supervised learning with synthetically generated paired data. However, obtaining high-quality paired data in real-world scenarios is often difficult. This data scarcity can degrade model performance under unseen conditions and limit generalization ability. To this end, in this work, we approach this problem from an unsupervised perspective, framing it as a probabilistic inverse problem. Our method requires only diffusion priors trained on individual sources. Separation is then achieved by iteratively guiding an initial state toward the solution through reconstruction guidance. Importantly, we introduce an advanced inverse problem solver specifically designed for separation, which mitigates gradient conflicts caused by interference between the diffusion prior and reconstruction guidance during inverse denoising. This design ensures high-quality and balanced separation performance across individual sources. Additionally, we find that initializing the denoising process with an augmented mixture instead of pure Gaussian noise provides an informative starting point that significantly improves the final performance. To further enhance audio prior modeling, we design a novel time-frequency attention-based network architecture that demonstrates strong audio modeling capability. Collectively, these improvements lead to significant performance gains, as validated across speech-sound event, sound event, and speech separation tasks.

eess.AS↗

arXiv2025

Grad: Guided Relation Diffusion Generation for Graph Augmentation in Graph Fraud Detection

Nowadays, Graph Fraud Detection (GFD) in financial scenarios has become an urgent research topic to protect online payment security. However, as organized crime groups are becoming more professional in real-world scenarios, fraudsters are employing more sophisticated camouflage strategies. Specifically, fraudsters disguise themselves by mimicking the behavioral data collected by platforms, ensuring that their key characteristics are consistent with those of benign users to a high degree, which we call Adaptive Camouflage. Consequently, this narrows the differences in behavioral traits between them and benign users within the platform's database, thereby making current GFD models lose efficiency. To address this problem, we propose a relation diffusion-based graph augmentation model Grad. In detail, Grad leverages a supervised graph contrastive learning module to enhance the fraud-benign difference and employs a guided relation diffusion generator to generate auxiliary homophilic relations from scratch. Based on these, weak fraudulent signals would be enhanced during the aggregation process, thus being obvious enough to be captured. Extensive experiments have been conducted on two real-world datasets provided by WeChat Pay, one of the largest online payment platforms with billions of users, and three public datasets. The results show that our proposed model Grad outperforms SOTA methods in both various scenarios, achieving at most 11.10% and 43.95% increases in AUC and AP, respectively. Our code is released at https://github.com/AI4Risk/antifraud and https://github.com/Muyiiiii/WWW25-Grad.

cs.LG↗

arXiv2025

Brain-Muscle Atlas: A novel framework for Motor Brain-Computer Interfaces

Motor brain-computer interfaces (BCIs) enable the control of external devices by decoding neural signals. However, most existing systems rely on a direct "brain-machine" mapping, overlooking the hierarchical physiological pathway of natural movement, namely the "brain-muscle-joint" cascade. Due to the lack of explicit modeling and enhancement of this pathway, current systems are often constrained by the low amplitude and high noise of EEG signals, resulting in motor outputs that are unstable, discontinuous, and insufficiently natural.To address these limitations, this study introduces the concept of a brain-muscle atlas, designed to systematically characterize the mapping between motor cortical activity and corresponding muscle activation, thereby establishing a movement decoding framework that better aligns with neuromuscular physiology. Using synchronously recorded EEG-EMG data, we constructed the first brain-muscle atlas for elbow flexion-extension, achieving a structured mapping from cortical activity to muscle activation.Offline experiments demonstrate that the proposed atlas accurately reconstructs the temporal activation patterns of primary elbow agonists, achieving a maximum correlation coefficient of 0.8314, thereby validating its ability to capture cortical-muscular mapping. Furthermore, by leveraging atlas-derived muscle activation representations, we enabled continuous real-time control of a virtual elbow joint. All ten participants successfully completed the online flexion-extension task, indicating that the system robustly extracts motor intent even under low-SNR EEG conditions.

cs.HC↗

arXiv2025

EVICPRESS: Joint KV-Cache Compression and Eviction for Efficient LLM Serving

Reusing KV cache is essential for high efficiency of Large Language Model (LLM) inference systems. With more LLM users, the KV cache footprint can easily exceed GPU memory capacity, so prior work has proposed to either evict KV cache to lower-tier storage devices, or compress KV cache so that more KV cache can be fit in the fast memory. However, prior work misses an important opportunity: jointly optimizing the eviction and compression decisions across all KV caches to minimize average generation latency without hurting quality. We propose EVICPRESS, a KV-cache management system that applies lossy compression and adaptive eviction to KV cache across multiple storage tiers. Specifically, for each KV cache of a context, EVICPRESS considers the effect of compression and eviction of the KV cache on the average generation quality and delay across all contexts as a whole. To achieve this, EVICPRESS proposes a unified utility function that quantifies the effect of quality and delay of the lossy compression or eviction. To this end, EVICPRESS's profiling module periodically updates the utility function scores on all possible eviction-compression configurations for all contexts and places KV caches using a fast heuristic to rearrange KV caches on all storage tiers, with the goal of maximizing the utility function scores on each storage tier. Compared to the baselines that evict KV cache or compress KV cache, EVICPRESS achieves higher KV-cache hit rates on fast devices, i.e., lower delay, while preserving high generation quality by applying conservative compression to contexts that are sensitive to compression errors. Evaluation on 12 datasets and 5 models demonstrates that EVICPRESS achieves up to 2.19x faster time-to-first-token (TTFT) at equivalent generation quality.

cs.OS↗

arXiv2025

PIMRL: Physics-Informed Multi-Scale Recurrent Learning for Burst-Sampled Spatiotemporal Dynamics

Deep learning has shown strong potential in modeling complex spatiotemporal dynamics. However, most existing methods depend on densely and uniformly sampled data, which is often unavailable in practice due to sensor and cost limitations. In many real-world settings, such as mobile sensing and physical experiments, data are burst-sampled with short high-frequency segments followed by long gaps, making it difficult to learn accurate dynamics from sparse observations. To address this issue, we propose Physics-Informed Multi-Scale Recurrent Learning (PIMRL), a novel framework specifically designed for burst-sampled spatiotemporal data. PIMRL combines macro-scale latent dynamics inference with micro-scale adaptive refinement guided by incomplete prior information from partial differential equations (PDEs). It further introduces a temporal message-passing mechanism to effectively propagate information across burst intervals. This multi-scale architecture enables PIMRL to model complex systems accurately even under severe data scarcity. We evaluate our approach on five benchmark datasets involving 1D to 3D multi-scale PDEs. The results show that PIMRL consistently outperforms state-of-the-art baselines, achieving substantial improvements and reducing errors by up to 80% in the most challenging settings, which demonstrates the clear advantage of our model. Our work demonstrates the effectiveness of physics-informed recurrent learning for accurate and efficient modeling of sparse spatiotemporal systems.

cs.LG↗

arXiv2025

Stable spectral neural operator for learning stiff PDE systems from limited data

Accurate modeling of spatiotemporal dynamics is crucial to understanding complex phenomena across science and engineering. However, this task faces a fundamental challenge when the governing equations are unknown and observational data are sparse. System stiffness, the coupling of multiple time-scales, further exacerbates this problem and hinders long-term prediction. Existing methods fall short: purely data-driven methods demand massive datasets, whereas physics-aware approaches are constrained by their reliance on known equations and fine-grained time steps. To overcome these limitations, we introduce an equation-free learning framework, namely, the Stable Spectral Neural Operator (SSNO), for modeling stiff partial differential equation (PDE) systems based on limited data. Instead of encoding specific equation terms, SSNO embeds spectrally inspired structures in its architecture, yielding strong inductive biases for learning the underlying physics. It automatically learns local and global spatial interactions in the frequency domain, while handling system stiffness with a robust integrating factor time-stepping scheme. Demonstrated across multiple 2D and 3D benchmarks in Cartesian and spherical geometries, SSNO achieves prediction errors one to two orders of magnitude lower than leading models. Crucially, it shows remarkable data efficiency, requiring only very few (2--5) training trajectories for robust generalization to out-of-distribution conditions. This work offers a robust and generalizable approach to learning stiff spatiotemporal dynamics from limited data without explicit \textit{a priori} knowledge of PDE terms.

physics.comp-ph↗

arXiv2025

QiMeng-SALV: Signal-Aware Learning for Verilog Code Generation

The remarkable progress of Large Language Models (LLMs) presents promising opportunities for Verilog code generation which is significantly important for automated circuit design. The lacking of meaningful functional rewards hinders the preference optimization based on Reinforcement Learning (RL) for producing functionally correct Verilog code. In this paper, we propose Signal-Aware Learning for Verilog code generation (QiMeng-SALV) by leveraging code segments of functionally correct output signal to optimize RL training. Considering Verilog code specifies the structural interconnection of hardware gates and wires so that different output signals are independent, the key insight of QiMeng-SALV is to extract verified signal-aware implementations in partially incorrect modules, so as to enhance the extraction of meaningful functional rewards. Roughly, we verify the functional correctness of signals in generated module by comparing with that of reference module in the training data. Then abstract syntax tree (AST) is employed to identify signal-aware code segments which can provide meaningful functional rewards from erroneous modules. Finally, we introduce signal-aware DPO which is optimized on the correct signal-level code segments, thereby preventing noise and interference from incorrect signals. The proposed QiMeng-SALV underscores the paradigm shift from conventional module-level to fine-grained signal-level optimization in Verilog code generation, addressing the issue of insufficient functional rewards. Experiments demonstrate that our method achieves state-of-the-art performance on VerilogEval and RTLLM, with a 7B parameter model matching the performance of the DeepSeek v3 671B model and significantly outperforming the leading open-source model CodeV trained on the same dataset. Our code is available at https://github.com/QiMeng-IPRC/QiMeng-SALV.

cs.LG↗

arXiv2025

LMCache: An Efficient KV Cache Layer for Enterprise-Scale LLM Inference

KV cache has traditionally been stored in GPU memory to accelerate the decoding phase of large language model (LLM) inference. However, it is increasingly necessary to move KV caches outside GPU devices, to enable cache reuse across different queries and inference engines. Our real-world usage statistics confirm this trend: over time, the total KV cache stored by users has grown rapidly, far exceeding the capacity of GPU memory. Despite this need, there lacks an efficient solution for offloading and transferring KV caches. We present LMCACHE, the first and so far the most efficient open-source KV caching solution, which extracts and stores KV caches generated by modern LLM engines (vLLM and SGLang) out of the GPU memory and shares them across engines and queries. LMCACHE supports both cache offloading (prefix reuse across queries) and prefill-decode (PD) disaggregation (cross-engine/GPU cache transfer). LMCACHE's high performance and wide adoption stem from the following contributions: (1) highly optimized KV cache data movement powered by batched data movement operations, compute and I/O pipelining; (2) a modular KV cache connector component, decoupling LMCACHE from the rapid evolution of inference engines; (3) a first-class control API for flexible cache orchestration across GPU, CPU, storage, and network layers. Our evaluation shows that combining LMCACHE with vLLM achieves up to 15x improvement in throughput across workloads such as multi-round question answering and document analysis. Large-scale adoption of LMCACHE in enterprise settings provides us valuable insights, for example, fetching KV cache from remote storage has unsurprisingly benefits to prefill delay, and that context truncation, which is a widely applied technique in industry, can greatly reduce prefix cache hit ratio by half. The source code of LMCACHE is at: https://github.com/LMCache/LMCache.

cs.LG↗

arXiv2025

Likelihood-free Posterior Density Learning for Uncertainty Quantification in Inference Problems

Generative models and those with computationally intractable likelihoods are widely used to describe complex systems in the natural sciences, social sciences, and engineering. Fitting these models to data requires likelihood-free inference methods that explore the parameter space without explicit likelihood evaluations, relying instead on sequential simulation, which comes at the cost of computational efficiency and extensive tuning. We develop an alternative framework called kernel-adaptive synthetic posterior estimation (KASPE) that uses deep learning to directly reconstruct the mapping between the observed data and a finite-dimensional parametric representation of the posterior distribution, trained on a large number of simulated datasets. We provide theoretical justification for KASPE and a formal connection to the likelihood-based approach of expectation propagation. Simulation experiments demonstrate KASPE's flexibility and performance relative to existing likelihood-free methods including approximate Bayesian computation in challenging inferential settings involving posteriors with heavy tails, multiple local modes, and over the parameters of a nonlinear dynamical system.

stat.ME↗

arXiv2025

Universal Sensitivity Bound for Thermal Quantum Dynamic Sensing

This work unifies the equilibrium and non-equilibrium frameworks of quantum metrology within the context of many-body systems. We investigate dynamic sensing schemes to derive an upper bound on the quantum Fisher information for probe states in thermal equilibrium with their environment. We establish that the dynamic quantum Fisher information for a thermal probe state is upper bounded by the degree of non-commutation between the transformed local generator and the Hamiltonian for the thermal state. Furthermore, we show that this upper bound scales as the square of the product of the inverse temperature and the evolution time. In the low-temperature limit, we establish an additional upper bound expressed as the seminorm of the commutator divided by the energy gap. We apply this thermal dynamic sensing scheme to various models, demonstrating that the dynamic quantum Fisher information satisfies the established upper bounds.

quant-ph↗

arXiv2025

DEAL-300K: Diffusion-based Editing Area Localization with a 300K-Scale Dataset and Frequency-Prompted Baseline

Diffusion-based image editing has made semantic level image manipulation easy for general users, but it also enables realistic local forgeries that are hard to localize. Existing benchmarks mainly focus on the binary detection of generated images or the localization of manually edited regions and do not reflect the properties of diffusion-based edits, which often blend smoothly into the original content. We present Diffusion-Based Image Editing Area Localization Dataset (DEAL-300K), a large scale dataset for diffusion-based image manipulation localization (DIML) with more than 300,000 annotated images. We build DEAL-300K by using a multi-modal large language model to generate editing instructions, a mask-free diffusion editor to produce manipulated images, and an active-learning change detection pipeline to obtain pixel-level annotations. On top of this dataset, we propose a localization framework that uses a frozen Visual Foundation Model (VFM) together with Multi Frequency Prompt Tuning (MFPT) to capture both semantic and frequency-domain cues of edited regions. Trained on DEAL-300K, our method reaches a pixel-level F1 score of 82.56% on our test split and 80.97% on the external CoCoGlide benchmark, providing strong baselines and a practical foundation for future DIML research.The dataset can be accessed via https://github.com/ymhzyj/DEAL-300K.

cs.CV↗

arXiv2025

Efficient Diffusion Planning with Temporal Diffusion

Diffusion planning is a promising method for learning high-performance policies from offline data. To avoid the impact of discrepancies between planning and reality on performance, previous works generate new plans at each time step. However, this incurs significant computational overhead and leads to lower decision frequencies, and frequent plan switching may also affect performance. In contrast, humans might create detailed short-term plans and more general, sometimes vague, long-term plans, and adjust them over time. Inspired by this, we propose the Temporal Diffusion Planner (TDP) which improves decision efficiency by distributing the denoising steps across the time dimension. TDP begins by generating an initial plan that becomes progressively more vague over time. At each subsequent time step, rather than generating an entirely new plan, TDP updates the previous one with a small number of denoising steps. This reduces the average number of denoising steps, improving decision efficiency. Additionally, we introduce an automated replanning mechanism to prevent significant deviations between the plan and reality. Experiments on D4RL show that, compared to previous works that generate new plans every time step, TDP improves the decision-making frequency by 11-24.8 times while achieving higher or comparable performance.

cs.LG↗

arXiv2025

Movable-Antenna Array Enhanced Multi-Target Sensing: CRB Characterization and Optimization

Movable antennas (MAs) have emerged as a promising technology to improve wireless communication and sensing performance towards sixth-generation (6G) networks through flexible antenna movement. In this paper, we propose a novel wireless sensing system based on MA arrays to enhance multi-target spatial angle estimation performance. We begin by characterizing the Cramér-Rao bound (CRB) matrix for multi-target angle of arrival (AoA) estimation as a function of the antenna's positions in MA arrays, thereby establishing a theoretical foundation for antenna position optimization. Then, aiming at improving the sensing coverage performance, we formulate an optimization problem to minimize the expectation of the trace of the CRB matrix over random target angles subject to a given distribution by optimizing the antennas' positions. To tackle the formulated challenging optimization problem, the Monte Carlo method is employed to approximate the intractable objective function, and a swarm-based gradient descent algorithm is subsequently proposed to address the approximated problem. In addition, a lower-bound on the sum of CRBs for multi-target AoA estimation is derived. Numerical results demonstrate that the proposed MA-based design achieves superior sensing performance compared to conventional systems using fixed-position antenna (FPA) arrays and single-target-oriented MA arrays, in terms of decreasing both CRB and the actual AoA estimation mean square error (MSE). Fundamentally, the designed MA array geometry exhibits low correlation and high effective power of sensitivity vectors for multi-target sensing in the angular domain, leading to significant CRB performance improvement. The resultant low correlation of steering vectors over multiple targets' directions further helps mitigate angle estimation ambiguity and thus enhances MSE performance.

eess.SP↗

arXiv2025

MammothModa2: A Unified AR-Diffusion Framework for Multimodal Understanding and Generation

Unified multimodal models aim to integrate understanding and generation within a single framework, yet bridging the gap between discrete semantic reasoning and high-fidelity visual synthesis remains challenging. We present MammothModa2 (Mammoth2), a unified autoregressive-diffusion (AR-Diffusion) framework designed to effectively couple autoregressive semantic planning with diffusion-based generation. Mammoth2 adopts a serial design: an AR path equipped with generation experts performs global semantic modeling over discrete tokens, while a single-stream Diffusion Transformer (DiT) decoder handles high-fidelity image synthesis. A carefully designed AR-Diffusion feature alignment module combines multi-layer feature aggregation, unified condition encoding, and in-context conditioning to stably align AR's representations with the diffusion decoder's continuous latents. Mammoth2 is trained end-to-end with joint Next-Token Prediction and Flow Matching objectives, followed by supervised fine-tuning and reinforcement learning over both generation and editing. With roughly 60M supervised generation samples and no reliance on pre-trained generators, Mammoth2 delivers strong text-to-image and instruction-based editing performance on public benchmarks, achieving 0.87 on GenEval, 87.2 on DPGBench, and 4.06 on ImgEdit, while remaining competitive with understanding-only backbones (e.g., Qwen3-VL-8B) on multimodal understanding tasks. These results suggest that a carefully coupled AR-Diffusion architecture can provide high-fidelity generation and editing while maintaining strong multimodal comprehension within a single, parameter- and data-efficient model.

cs.CV↗