arXiv subjects

Zhi Chen

Publications and source records attributed to Zhi Chen.

At least 109 records · Page 6Linked to original sources

arXiv2025

Compressive Near-Field Wideband Channel Estimation for THz Extremely Large-scale MIMO Systems

We consider the channel acquisition problem for a wideband terahertz (THz) communication system, where an extremely large-scale array is deployed to mitigate severe path attenuation. In channel modeling, we account for both the near-field spherical wavefront and the wideband beam-splitting phenomena, resulting in a wideband near-field channel. We propose a frequency-independent orthogonal dictionary that generalizes the standard discrete Fourier transform (DFT) matrix by introducing an additional parameter to capture the near-field property. This dictionary enables the wideband near-field channel to be efficiently represented with a two-dimensional (2D) block-sparse structure. Leveraging this specific sparse structure, the wideband near-field channel estimation problem can be effectively addressed within a customized compressive sensing framework. Numerical results demonstrate the significant advantages of our proposed 2D block-sparsity-aware method over conventional polar-domain-based approaches for near-field wideband channel estimation.

eess.SP↗

arXiv2025

URPO: A Unified Reward & Policy Optimization Framework for Large Language Models

Large-scale alignment pipelines typically pair a policy model with a separately trained reward model whose parameters remain frozen during reinforcement learning (RL). This separation creates a complex, resource-intensive pipeline and suffers from a performance ceiling due to a static reward signal. We propose a novel framework, Unified Reward & Policy Optimization (URPO), that unifies instruction-following ("player") and reward modeling ("referee") within a single model and a single training phase. Our method recasts all alignment data-including preference pairs, verifiable reasoning, and open-ended instructions-into a unified generative format optimized by a single Group-Relative Policy Optimization (GRPO) loop. This enables the model to learn from ground-truth preferences and verifiable logic while simultaneously generating its own rewards for open-ended tasks. Experiments on the Qwen2.5-7B model demonstrate URPO's superiority. Our unified model significantly outperforms a strong baseline using a separate generative reward model, boosting the instruction-following score on AlpacaEval from 42.24 to 44.84 and the composite reasoning average from 32.66 to 35.66. Furthermore, URPO cultivates a superior internal evaluator as a byproduct of training, achieving a RewardBench score of 85.15 and surpassing the dedicated reward model it replaces (83.55). By eliminating the need for a separate reward model and fostering a co-evolutionary dynamic between generation and evaluation, URPO presents a simpler, more efficient, and more effective path towards robustly aligned language models.

cs.CV↗

arXiv2025

SVIP: Semantically Contextualized Visual Patches for Zero-Shot Learning

Zero-shot learning (ZSL) aims to recognize unseen classes without labeled training examples by leveraging class-level semantic descriptors such as attributes. A fundamental challenge in ZSL is semantic misalignment, where semantic-unrelated information involved in visual features introduce ambiguity to visual-semantic interaction. Unlike existing methods that suppress semantic-unrelated information post hoc either in the feature space or the model space, we propose addressing this issue at the input stage, preventing semantic-unrelated patches from propagating through the network. To this end, we introduce Semantically contextualized VIsual Patches (SVIP) for ZSL, a transformer-based framework designed to enhance visual-semantic alignment. Specifically, we propose a self-supervised patch selection mechanism that preemptively learns to identify semantic-unrelated patches in the input space. This is trained with the supervision from aggregated attention scores across all transformer layers, which estimate each patch's semantic score. As removing semantic-unrelated patches from the input sequence may disrupt object structure, we replace them with learnable patch embeddings. With initialization from word embeddings, we can ensure they remain semantically meaningful throughout feature extraction. Extensive experiments on ZSL benchmarks demonstrate that SVIP achieves state-of-the-art performance results while providing more interpretable and semantically rich feature representations. Code is available at https://github.com/uqzhichen/SVIP.

cs.CV↗

arXiv2025

Movable-Antenna-Enhanced Physical-Layer Service Integration: Performance Analysis and Optimization

Movable antennas (MAs) have drawn increasing attention in wireless communications due to their capability to create favorable channel conditions via local movement within a confined region. In this letter, we investigate its application in physical-layer service integration (PHY-SI), where a multi-MA base station (BS) simultaneously transmits both confidential and multicast messages to two users. The multicast message is intended for both users, while the confidential message is intended only for one user and must remain perfectly secure from the other. Our goal is to jointly optimize the secrecy and multicast beamforming, as well as the MAs' positions at the BS to maximize the secrecy rate for one user while satisfying the multicast rate requirement for both users. To gain insights, we first conduct performance analysis of this MA-enhanced PHY-SI system in two special cases, revealing its unique characteristics compared to conventional PHY-SI with fixed-position antennas (FPAs). To address the secrecy rate maximization problem, we propose a two-layer optimization framework that integrates the semidefinite relaxation (SDR) technique and a discrete sampling algorithm. Numerical results demonstrate that MAs can greatly enhance the achievable secrecy rate region for PHY-SI compared to FPAs.

cs.IT↗

arXiv2025

Movable Antennas Meet Intelligent Reflecting Surface: Friends or Foes?

Movable antenna (MA) and intelligent reflecting surface (IRS) are considered promising technologies for the next-generation wireless communication systems due to their shared channel reconfiguration capabilities. This, however, raises a fundamental question: Does the performance gain of MAs over conventional fixed-position antennas (FPAs) still exist in the presence of the IRS? To answer this question, we investigate in this paper an IRS-assisted multi-user multiple-input single-output (MISO) MA system, where a multi-MA base station (BS) transmits to multiple single-FPA users. We formulate a sum-rate maximization problem by jointly optimizing the active/passive beamforming of the BS/IRS and the MA positions within a one-dimensional transmit region, which is challenging to be optimally solved. To drive essential insights, we first study a simplified case with a single user. Then, we analyze the performance gain of MAs over FPAs in the light-of-sight (LoS) BS-IRS channel and derive the conditions under which this gain becomes more or less significant. In addition, we propose an alternating optimization (AO) algorithm to solve the signal-to-noise ratio (SNR) maximization problem in the single-user case by combining the block coordinate descent (BCD) method and the graph-based method. For the general multi-user case, our performance analysis unveils that the performance gain of MAs over FPAs diminishes with typical transmit precoding strategies at the BS under certain conditions. We also propose a high-quality suboptimal solution to the sum-rate maximization problem by applying the AO algorithm that combines the weighted minimum mean square error (WMMSE) algorithm, manifold optimization method and discrete sampling method. Numerical results validate our theoretical analyses and demonstrate that the performance gain of MAs over FPAs may be reduced if the IRS passive beamforming is optimized.

cs.IT↗

arXiv2025

Derivative-Free Optimization-Empowered Wireless Channel Reconfiguration for 6G

Reconfigurable antennas, including reconfigurable intelligent surface (RIS), movable antenna (MA), fluid antenna (FA), and other advanced antenna techniques, have been studied extensively in the context of reshaping wireless propagation environments for 6G and beyond wireless communications. Nevertheless, how to reconfigure/optimize the real-time controllable coefficients to achieve a favorable end-to-end wireless channel remains a substantial challenge, as it usually requires accurate modeling of the complex interaction between the reconfigurable devices and the electromagnetic waves, as well as knowledge of implicit channel propagation parameters. In this paper, we introduce a derivative-free optimization (a.k.a., zeroth-order (ZO) optimization) technique to directly optimize reconfigurable coefficients to shape the wireless end-to-end channel, without the need of channel modeling and estimation of the implicit environmental propagation parameters. We present the fundamental principles of ZO optimization and discuss its potential advantages in wireless channel reconfiguration. Two case studies for RIS and movable antenna-enabled single-input single-output (SISO) systems are provided to show the superiority of ZO-based methods as compared to state-of-the-art techniques. Finally, we outline promising future research directions and offer concluding insights on derivative-free optimization for reconfigurable antenna technologies.

eess.SP↗

arXiv2025

Are Synthetic Videos Useful? A Benchmark for Retrieval-Centric Evaluation of Synthetic Videos

Text-to-video (T2V) synthesis has advanced rapidly, yet current evaluation metrics primarily capture visual quality and temporal consistency, offering limited insight into how synthetic videos perform in downstream tasks such as text-to-video retrieval (TVR). In this work, we introduce SynTVA, a new dataset and benchmark designed to evaluate the utility of synthetic videos for building retrieval models. Based on 800 diverse user queries derived from MSRVTT training split, we generate synthetic videos using state-of-the-art T2V models and annotate each video-text pair along four key semantic alignment dimensions: Object \& Scene, Action, Attribute, and Prompt Fidelity. Our evaluation framework correlates general video quality assessment (VQA) metrics with these alignment scores, and examines their predictive power for downstream TVR performance. To explore pathways of scaling up, we further develop an Auto-Evaluator to estimate alignment quality from existing metrics. Beyond benchmarking, our results show that SynTVA is a valuable asset for dataset augmentation, enabling the selection of high-utility synthetic samples that measurably improve TVR outcomes. Project page and dataset can be found at https://jasoncodemaker.github.io/SynTVA/.

cs.CV↗

arXiv2025

Round Attention: A Novel Round-Level Attention Mechanism to Accelerate LLM Inference

The increasing context window size in large language models (LLMs) has improved their ability to handle complex, long-text tasks. However, as the conversation rounds continue, it is required to store a large amount of KV cache in GPU memory, which significantly affects the efficiency and even availability of the model serving systems. This paper analyzes dialogue data from real users on the granularity of round and discovers that the LLM inference manifests a watershed layer, after which the distribution of round-level attention shows notable similarity. Based on this, we propose Round Attention - a novel round-level attention mechanism that selectively processes the KV cache of top-k relevant rounds, where k is dynamically determined through the attention matrix in the watershed layer. Theoretical analysis demonstrates that our method reduces memory usage by 54\% to 82\%, while experimental results confirm that loading sparse critical-round KV cache maintains answer accuracy without performance degradation.

cs.CL↗

arXiv2025

Capability Salience Vector: Fine-grained Alignment of Loss and Capabilities for Downstream Task Scaling Law

Scaling law builds the relationship between training computation and validation loss, enabling researchers to effectively predict the loss trending of models across different levels of computation. However, a gap still remains between validation loss and the model's downstream capabilities, making it untrivial to apply scaling law to direct performance prediction for downstream tasks. The loss typically represents a cumulative penalty for predicted tokens, which are implicitly considered to have equal importance. Nevertheless, our studies have shown evidence that when considering different training data distributions, we cannot directly model the relationship between downstream capability and computation or token loss. To bridge the gap between validation loss and downstream task capabilities, in this work, we introduce Capability Salience Vector, which decomposes the overall loss and assigns different importance weights to tokens to assess a specific meta-capability, aligning the validation loss with downstream task performance in terms of the model's capabilities. Experiments on various popular benchmarks demonstrate that our proposed Capability Salience Vector could significantly improve the predictability of language model performance on downstream tasks.

cs.CL↗

arXiv2025

FlowOE: Imitation Learning with Flow Policy from Ensemble RL Experts for Optimal Execution under Heston Volatility and Concave Market Impacts

Optimal execution in financial markets refers to the process of strategically transacting a large volume of assets over a period to achieve the best possible outcome by balancing the trade-off between market impact costs and timing or volatility risks. Traditional optimal execution strategies, such as static Almgren-Chriss models, often prove suboptimal in dynamic financial markets. This paper propose flowOE, a novel imitation learning framework based on flow matching models, to address these limitations. FlowOE learns from a diverse set of expert traditional strategies and adaptively selects the most suitable expert behavior for prevailing market conditions. A key innovation is the incorporation of a refining loss function during the imitation process, enabling flowOE not only to mimic but also to improve upon the learned expert actions. To the best of our knowledge, this work is the first to apply flow matching models in a stochastic optimal execution problem. Empirical evaluations across various market conditions demonstrate that flowOE significantly outperforms both the specifically calibrated expert models and other traditional benchmarks, achieving higher profits with reduced risk. These results underscore the practical applicability and potential of flowOE to enhance adaptive optimal execution.

cs.LG↗

arXiv2025

VModA: An Effective Framework for Adaptive NSFW Image Moderation

Not Safe/Suitable for Work (NSFW) content is rampant on social networks and poses serious harm to citizens, especially minors. Current detection methods mainly rely on deep learning-based image recognition and classification. However, NSFW images are now presented in increasingly sophisticated ways, often using image details and complex semantics to obscure their true nature or attract more views. Although still understandable to humans, these images often evade existing detection methods, posing a significant threat. Further complicating the issue, varying regulations across platforms and regions create additional challenges for effective moderation, leading to detection bias and reduced accuracy. To address this, we propose VModA, a general and effective framework that adapts to diverse moderation rules and handles complex, semantically rich NSFW content across categories. Experimental results show that VModA significantly outperforms existing methods, achieving up to a 54.3% accuracy improvement across NSFW types, including those with complex semantics. Further experiments demonstrate that our method exhibits strong adaptability across categories, scenarios, and base VLMs. We also identified inconsistent and controversial label samples in public NSFW benchmark datasets, re-annotated them, and submitted corrections to the original maintainers. Two datasets have confirmed the updates so far. Additionally, we evaluate VModA in real-world scenarios to demonstrate its practical effectiveness.

cs.CV↗

arXiv2025

CNN-Based Channel Map Estimation for Movable Antenna Systems

Movable antenna (MA) has attracted increasing attention in wireless communications due to its capability of wireless channel reconfiguration through local antenna movement within a confined region at the transmitter/receiver. However, to determine the optimal antenna positions, channel state information (CSI) within the entire region, termed small-scale channel map, is required, which poses a significant challenge due to the unaffordable overhead for exhaustive channel estimation at all positions. To tackle this challenge, in this paper, we propose a new convolutional neural network (CNN)-based estimation scheme to reconstruct the small-scale channel map within a three-dimensional (3D) movement region. Specifically, we first collect a set of CSI measurements corresponding to a subset of MA positions and different receiver locations offline to comprehensively capture the environmental features. Subsequently, we train a CNN using the collected data, which is then used to reconstruct the full channel map during real-time transmission only based on a finite number of channel measurements taken at several selected MA positions within the 3D movement region. Numerical results demonstrate that our proposed scheme can accurately reconstruct the small-scale channel map and outperforms other benchmark schemes.

eess.SP↗

arXiv2025

FlowHFT: Imitation Learning via Flow Matching Policy for Optimal High-Frequency Trading under Diverse Market Conditions

High-frequency trading (HFT) is an investing strategy that continuously monitors market states and places bid and ask orders at millisecond speeds. Traditional HFT approaches fit models with historical data and assume that future market states follow similar patterns. This limits the effectiveness of any single model to the specific conditions it was trained for. Additionally, these models achieve optimal solutions only under specific market conditions, such as assumptions about stock price's stochastic process, stable order flow, and the absence of sudden volatility. Real-world markets, however, are dynamic, diverse, and frequently volatile. To address these challenges, we propose the FlowHFT, a novel imitation learning framework based on flow matching policy. FlowHFT simultaneously learns strategies from numerous expert models, each proficient in particular market scenarios. As a result, our framework can adaptively adjust investment decisions according to the prevailing market state. Furthermore, FlowHFT incorporates a grid-search fine-tuning mechanism. This allows it to refine strategies and achieve superior performance even in complex or extreme market scenarios where expert strategies may be suboptimal. We test FlowHFT in multiple market environments. We first show that flow matching policy is applicable in stochastic market environments, thus enabling FlowHFT to learn trading strategies under different market conditions. Notably, our single framework consistently achieves performance superior to the best expert for each market condition.

q-fin.TR↗

arXiv2025

STAR-RIS-Enabled Multi-Path Beam Routing with Passive Beam Splitting

Reconfigurable intelligent surfaces (RISs) can be densely deployed in the environment to create multi-reflection line-of-sight (LoS) links for signal coverage enhancement. However, conventional reflection-only RISs can only achieve half-space reflection, which limits the LoS path diversity. In contrast, simultaneously transmitting and reflecting RISs (STAR-RISs) can achieve full-space reflection and transmission, thereby creating more LoS paths. Hence, in this paper, we study a new multi-STAR-RIS-aided communication system, where a multi-antenna base station (BS) transmits to multiple single-antenna users by exploiting the signal beam routing over a set of cascaded LoS paths each formed by multiple STAR-RISs. To reveal essential insights, we first consider a simplified single-user case, aiming to maximize its received signal power by jointly optimizing the active beamforming at the BS, the BS's power allocation over different paths, the number of selected beam-routing paths, the selected STAR-RISs for each path, as well as their amplitude and phase shifts for transmission/reflection. However, this problem is difficult to be optimally solved as different paths may be intricately coupled at their shared STAR-RISs. To tackle this difficulty, we first derive the optimal solution to this problem in closed-form for a given set of paths. The clique-based approach in graph theory is then applied to solve the remaining multi-path selection problem efficiently. Next, we extend the proposed clique-based method to the multi-user case to maximize the minimum received signal power among all users, subject to additional constraints on the disjointness of the selected paths for different users. Simulation results show that our proposed STAR-RIS-enabled beam routing outperforms the conventional beam routing with reflection-only RISs in both single- and multi-user cases.

eess.SP↗

arXiv2025

What are the Essential Factors in Crafting Effective Long Context Multi-Hop Instruction Datasets? Insights and Best Practices

Recent advancements in large language models (LLMs) with extended context windows have significantly improved tasks such as information extraction, question answering, and complex planning scenarios. In order to achieve success in long context tasks, a large amount of work has been done to enhance the long context capabilities of the model through synthetic data. Existing methods typically utilize the Self-Instruct framework to generate instruction tuning data for better long context capability improvement. However, our preliminary experiments indicate that less than 35% of generated samples are multi-hop, and more than 40% exhibit poor quality, limiting comprehensive understanding and further research. To improve the quality of synthetic data, we propose the Multi-agent Interactive Multi-hop Generation (MIMG) framework, incorporating a Quality Verification Agent, a Single-hop Question Generation Agent, a Multiple Question Sampling Strategy, and a Multi-hop Question Merger Agent. This framework improves the data quality, with the proportion of high-quality, multi-hop, and diverse data exceeding 85%. Furthermore, we systematically investigate strategies for document selection, question merging, and validation techniques through extensive experiments across various models. Our findings show that our synthetic high-quality long-context instruction data significantly enhances model performance, even surpassing models trained on larger amounts of human-annotated data. Our code is available at: https://github.com/WowCZ/LongMIT.

cs.CL↗

arXiv2025

Task-Specific Data Selection for Instruction Tuning via Monosemantic Neuronal Activations

Instruction tuning improves the ability of large language models (LLMs) to follow diverse human instructions, but achieving strong performance on specific target tasks remains challenging. A critical bottleneck is selecting the most relevant data to maximize task-specific performance. Existing data selection approaches include unstable influence-based methods and more stable distribution alignment methods, the latter of which critically rely on the underlying sample representation. In practice, most distribution alignment methods, from shallow features (e.g., BM25) to neural embeddings (e.g., BGE, LLM2Vec), may fail to capture how the model internally processes samples. To bridge this gap, we adopt a model-centric strategy in which each sample is represented by its neuronal activation pattern in the model, directly reflecting internal computation. However, directly using raw neuron activations leads to spurious similarity between unrelated samples due to neuron polysemanticity, where a single neuron may respond to multiple, unrelated concepts. To address this, we employ sparse autoencoders to disentangle polysemantic activations into sparse, monosemantic representations, and introduce a dedicated similarity metric for this space to better identify task-relevant data. Comprehensive experiments across multiple instruction datasets, models, tasks, and selection ratios show that our approach consistently outperforms existing data selection baselines in both stability and task-specific performance.

cs.LG↗

arXiv2025

Robust Movable-Antenna Position Optimization with Imperfect CSI for MISO Systems

Movable antenna (MA) technology has emerged as a promising solution for reconfiguring wireless channel conditions through local antenna movement within confined regions. Unlike previous works assuming perfect channel state information (CSI), this letter addresses the robust MA position optimization problem under imperfect CSI conditions for a multiple-input single-output (MISO) MA system. Specifically, we consider two types of CSI errors: norm-bounded and randomly distributed errors, aiming to maximize the worst-case and non-outage received signal power, respectively. For norm-bounded CSI errors, we derive the worst-case received signal power in closed-form. For randomly distributed CSI errors, due to the intractability of the probabilistic constraints, we apply the Bernstein-type inequality to obtain a closed-form lower bound for the non-outage received signal power. Based on these results, we show the optimality of the maximum-ratio transmission for imperfect CSI in both scenarios and employ a graph-based algorithm to obtain the optimal MA positions. Numerical results show that our proposed scheme can even outperform other benchmark schemes implemented under perfect CSI conditions.

cs.IT↗

arXiv2025

Mechanical Power Modeling and Energy Efficiency Maximization for Movable Antenna Systems

Movable antennas (MAs) have recently garnered significant attention in wireless communications due to their capability to reshape wireless channels via local antenna movement within a confined region. However, to achieve accurate antenna movement, MA drivers introduce non-negligible mechanical power consumption, rendering energy efficiency (EE) optimization more critical compared to conventional fixed-position antenna (FPA) systems. To address this problem, we develop in this paper a fundamental power consumption model for stepper motor-driven MA systems by resorting to basic electric motor theory. Based on this model, we formulate an EE maximization problem by jointly optimizing an MA's position, moving speed, and transmit power. However, this problem is difficult to solve optimally due to the intricate relationship between the mechanical power consumption and the design variables. To tackle this issue, we first uncover a hidden monotonicity of the EE performance with respect to the MA's moving speed. Then, we apply the Dinkelbach algorithm to obtain the optimal transmit power in a semi-closed form for any given MA position, followed by an enumeration to determine the optimal MA position. Numerical results demonstrate that despite the additional mechanical power consumption, the MA system can outperform the conventional FPA system in terms of EE.

cs.IT↗