arXiv subjects

Fan Zhang

Publications and source records attributed to Fan Zhang.

At least 361 records · Page 20Linked to original sources

arXiv2025

Near-Field ISAC for THz Wireless Systems

Sixth-generation (6G) wireless networks are expected not only to provide high-speed connectivity but also to support reliable sensing capabilities, giving rise to the integrated sensing and communication (ISAC) paradigm. To enable higher data rates and more accurate sensing, terahertz (THz) systems empowered by extremely large multiple-input-multiple-output (XL-MIMO) technology are envisioned as key enablers for future ISAC systems. Owing to the substantial increase in both effective array aperture and carrier frequency, a considerable portion of future ISAC applications is anticipated to fall within the near-field coverage region, instead of the conventional far-field. However, most existing ISAC techniques are designed under the far-field planar wave assumption, struggling to accommodate the unique characteristics of THz near-field propagation. To motivate future research into near-field ISAC research, we systematically investigate the characteristics of THz near-field propagation and explore its potential to facilitate ISAC systems. Specifically, we analyze three fundamental characteristics of THz near-field propagation and review state-of-the-art techniques that exploit these features to boost both communication and sensing performance. To further harness the angular-range coupling effect, we zoom into a particularly interesting approach to near-field sensing based on wavenumber domain. Besides, to exploit the beam squint effect, an ISAC resource allocation framework is introduced to support integrated multi-angle sensing and multi-user communication. Finally, we outline promising directions for future research in this emerging area.

eess.SP↗

arXiv2025

Leveraging Multimodal Data and Side Users for Diffusion Cross-Domain Recommendation

Cross-domain recommendation (CDR) aims to address the persistent cold-start problem in Recommender Systems. Current CDR research concentrates on transferring cold-start users' information from the auxiliary domain to the target domain. However, these systems face two main issues: the underutilization of multimodal data, which hinders effective cross-domain alignment, and the neglect of side users who interact solely within the target domain, leading to inadequate learning of the target domain's vector space distribution. To address these issues, we propose a model leveraging Multimodal data and Side users for diffusion Cross-domain recommendation (MuSiC). We first employ a multimodal large language model to extract item multimodal features and leverage a large language model to uncover user features using prompt learning without fine-tuning. Secondly, we propose the cross-domain diffusion module to learn the generation of feature vectors in the target domain. This approach involves learning feature distribution from side users and understanding the patterns in cross-domain transformation through overlapping users. Subsequently, the trained diffusion module is used to generate feature vectors for cold-start users in the target domain, enabling the completion of cross-domain recommendation tasks. Finally, our experimental evaluation of the Amazon dataset confirms that MuSiC achieves state-of-the-art performance, significantly outperforming all selected baselines. Our code is available: https://anonymous.4open.science/r/MuSiC-310A/.

cs.IR↗

arXiv2025

GPAS: Accelerating Convergence of LLM Pretraining via Gradient-Preserving Activation Scaling

Modern Large Language Models, such as the LLaMA, Qwen and DeepSeek series, predominantly adopt the Pre-LayerNorm (Pre-LN) Transformer architecture. While being stable during pretraining and scalable to large model sizes, Pre-LN suffers from an exponential growth in activation variance across layers, causing the shortcut to dominate over sub-layer outputs in the residual connection and limiting the learning capacity of deeper layers. To mitigate this issue, we propose Gradient-Preserving Activation Scaling (GPAS), a simple technique that can be used in combination with existing approaches. GPAS works by scaling down the intermediate activations while keeping their gradients unchanged. This leaves information in the activations intact, and avoids the gradient vanishing problem associated with gradient downscaling. Extensive experiments across various model sizes from 71M to 1B show that GPAS achieves consistent performance gains. Beyond enhancing Pre-LN Transformers, GPAS also shows promise in improving alternative architectures such as Sandwich-LN and DeepNorm, demonstrating its versatility and potential for improving training dynamics in a wide range of settings. Our code is available at https://github.com/dandingsky/GPAS.

cs.LG↗

arXiv2025

ShotBench: Expert-Level Cinematic Understanding in Vision-Language Models

Cinematography, the fundamental visual language of film, is essential for conveying narrative, emotion, and aesthetic quality. While recent Vision-Language Models (VLMs) demonstrate strong general visual understanding, their proficiency in comprehending the nuanced cinematic grammar embedded within individual shots remains largely unexplored and lacks robust evaluation. This critical gap limits both fine-grained visual comprehension and the precision of AI-assisted video generation. To address this, we introduce ShotBench, a comprehensive benchmark specifically designed for cinematic language understanding. It features over 3.5k expert-annotated QA pairs from images and video clips, meticulously curated from over 200 acclaimed (predominantly Oscar-nominated) films and spanning eight key cinematography dimensions. Our evaluation of 24 leading VLMs on ShotBench reveals their substantial limitations: even the top-performing model achieves less than 60% average accuracy, particularly struggling with fine-grained visual cues and complex spatial reasoning. To catalyze advancement in this domain, we construct ShotQA, a large-scale multimodal dataset comprising approximately 70k cinematic QA pairs. Leveraging ShotQA, we develop ShotVL through supervised fine-tuning and Group Relative Policy Optimization. ShotVL significantly outperforms all existing open-source and proprietary models on ShotBench, establishing new state-of-the-art performance. We open-source our models, data, and code to foster rapid progress in this crucial area of AI-driven cinematic understanding and generation.

cs.CV↗

arXiv2025

Beijing Normal University 12-meter Interferometric kHz GW Detector Prototype: Design and Scientific Prospects

Current gravitational-wave detectors have achieved remarkable sensitivity around 100 Hz, enabling ground-breaking discoveries. Enhancing sensitivity at higher frequencies in the kilohertz (kHz) range promises access to rich physics, particularly the extreme conditions during the merger stage of binary neutron stars. However, the high-frequency sensitivity of Michelson-based interferometers is fundamentally limited by their linear optical cavities, which are optimized for low-frequency signal enhancement. In [Phys. Rev. X 13, 021019 (2023)], a new configuration employing an L-shaped optical resonator was proposed to overcome this limitation, offering exceptional sensitivity in the kHz band. As a pathfinder, the 12-meter prototype at Beijing Normal University is designed to demonstrate the sensing and control schemes of this new kHz detector configuration and to explore its performance in the high-power regime with suspended optics. Beyond its primary scientific goal, the prototype also offers potential sensitivity in the megahertz (MHz) range, potentially enabling constraints on exotic sources. This paper presents an overview of the prototype, including its optical design and current development status of key components.

physics.optics↗

arXiv2025

CARMA: Context-Aware Situational Grounding of Human-Robot Group Interactions by Combining Vision-Language Models with Object and Action Recognition

We introduce CARMA, a system for situational grounding in human-robot group interactions. Effective collaboration in such group settings requires situational awareness based on a consistent representation of present persons and objects coupled with an episodic abstraction of events regarding actors and manipulated objects. This calls for a clear and consistent assignment of instances, ensuring that robots correctly recognize and track actors, objects, and their interactions over time. To achieve this, CARMA uniquely identifies physical instances of such entities in the real world and organizes them into grounded triplets of actors, objects, and actions. To validate our approach, we conducted three experiments, where multiple humans and a robot interact: collaborative pouring, handovers, and sorting. These scenarios allow the assessment of the system's capabilities as to role distinction, multi-actor awareness, and consistent instance identification. Our experiments demonstrate that the system can reliably generate accurate actor-action-object triplets, providing a structured and robust foundation for applications requiring spatiotemporal reasoning and situated decision-making in collaborative settings.

cs.RO↗

arXiv2025

Cavity-Mediated Gas-Liquid Transition

We study the gas-liquid transition in a binary Bose-Einstein condensate, where the two Zeeman-shifted hyperfine spin components are coupled by cavity-assisted Raman processes. Below a critical Zeeman field, the cavity becomes superradiant for an infinitesimally small pumping strength, where the enhanced superradiance is facilitated by the simultaneous formation of quantum droplet, a self-bound liquid phase stabilized by quantum fluctuations. Above the critical Zeeman field, the gas-liquid transition only takes place at a finite pumping strength after the system becomes superradiant. As the back action of the gas-liquid transition, the superradiant cavity field undergoes an abrupt jump at the first-order transition point. Furthermore, as a result of the fixed density ratio of the quantum droplet, the cavity field exhibits a linear scaling with the pumping strength in the liquid phase. These features serve as prominent signals for the cavity-mediated gas-liquid transition and coexistence, which derive from the interplay of Zeeman field, cavity-assisted spin mixing, and quantum fluctuations.

cond-mat.quant-gas↗

arXiv2025

Compressed Video Super-Resolution based on Hierarchical Encoding

This paper presents a general-purpose video super-resolution (VSR) method, dubbed VSR-HE, specifically designed to enhance the perceptual quality of compressed content. Targeting scenarios characterized by heavy compression, the method upscales low-resolution videos by a ratio of four, from 180p to 720p or from 270p to 1080p. VSR-HE adopts hierarchical encoding transformer blocks and has been sophisticatedly optimized to eliminate a wide range of compression artifacts commonly introduced by H.265/HEVC encoding across various quantization parameter (QP) levels. To ensure robustness and generalization, the model is trained and evaluated under diverse compression settings, allowing it to effectively restore fine-grained details and preserve visual fidelity. The proposed VSR-HE has been officially submitted to the ICME 2025 Grand Challenge on VSR for Video Conferencing (Team BVI-VSR), under both the Track 1 (General-Purpose Real-World Video Content) and Track 2 (Talking Head Videos).

eess.IV↗

arXiv2025

A Hierarchical Constructive Heuristic for Large-Scale Survivable Traffic Grooming Problem under Double-Link Failures

This paper studies a survivable traffic grooming problem in large-scale optical transport networks under double-link failures (STG2). Each communication demand must be assigned a route for every possible scenario involving zero, one, or two failed fiber links. Protection against double-link failures is crucial for ensuring reliable telecommunications services while minimizing equipment costs, making it essential for telecommunications companies today. However, this significantly complicates the problem and is rarely addressed in existing studies. Furthermore, current research typically examines networks with fewer than 300 nodes, much smaller than some emerging networks containing thousands of nodes. To address these challenges, we propose a novel hierarchical constructive heuristic for STG2. This heuristic constructs and assigns routes to communication demands across different scenarios by following a hierarchical sequence. It incorporates several innovative optimization techniques and utilizes parallel computing to enhance efficiency. Extensive experiments have been conducted on large-scale STG2 instances provided by our industry partner, encompassing networks with 1,000 to 2,600 nodes. Results demonstrate that within a one-hour time limit and a 16 GB memory limit set by the industry partner, our heuristic improves the objective values of the best-known solutions by 18.5\% on average, highlighting its significant potential for practical applications.

math.CO↗

arXiv2025

LP Relaxations for Routing and Wavelength Assignment with Partial Path Protection: Formulations and Computations

As a variant of the routing and wavelength assignment problem (RWAP), the RWAP with partial path protection (RWAP-PPP) designs a reliable optical-fiber network for telecommunications. It assigns paths and wavelengths to meet communication requests, not only in normal working situations but also in potential failure cases where an optical link fails. The literature lacks efficient relaxations to produce tight lower bounds on the optimal objective value of the RWAP-PPP. Consequently, the solution quality for the RWAP-PPP cannot be properly assessed, which is critical for telecommunication providers in customer bidding and service improvement. Due to numerous failure scenarios, developing effective lower bounds for the RWAP-PPP is challenging. To address this, we formulate and analyze various linear programming (LP) relaxations of the RWAP-PPP. Among them, we propose a novel LP relaxation yielding promising lower bounds. To solve it, we develop a Benders decomposition algorithm with valid inequalities to enhance performance. Computational results on practical networks, including large ones with hundreds of nodes and edges, demonstrate the effectiveness of the LP relaxation and efficiency of its algorithm. The obtained lower bounds achieve average optimality gaps of 8.6%. Compared with a direct LP relaxation of the RWAP, which has average gaps of 36.7%, significant improvements are observed. Consequently, our LP relaxation and algorithm effectively assess RWAP-PPP solution quality, offering significant research and practical value.

math.OC↗

arXiv2025

Unifying Lane-Level Traffic Prediction from a Graph Structural Perspective: Benchmark and Baseline

Traffic prediction has long been a focal and pivotal area in research, witnessing both significant strides from city-level to road-level predictions in recent years. With the advancement of Vehicle-to-Everything (V2X) technologies, autonomous driving, and large-scale models in the traffic domain, lane-level traffic prediction has emerged as an indispensable direction. However, further progress in this field is hindered by the absence of comprehensive and unified evaluation standards, coupled with limited public availability of data and code. In this paper, we present the first systematic classification framework for lane-level traffic prediction, offering a structured taxonomy and analysis of existing methods. We construct three representative datasets from two real-world road networks, covering both regular and irregular lane configurations, and make them publicly available to support future research. We further establishes a unified spatial topology structure and prediction task formulation, and proposes a simple yet effective baseline model, GraphMLP, based on graph structure and MLP networks. This unified framework enables consistent evaluation across datasets and modeling paradigms. We also reproduce previously unavailable code from existing studies and conduct extensive experiments to assess a range of models in terms of accuracy, efficiency, and applicability, providing the first benchmark that jointly considers predictive performance and training cost for lane-level traffic scenarios. All datasets and code are released at https://github.com/ShuhaoLii/LaneLevel-Traffic-Benchmark.

cs.LG↗

arXiv2025

The Throughput Gain of Hypercycle-level Resource Reservation for Time-Triggered Ethernet

Time-Triggered Communication is a key technology for many safety-critical systems, with applications spanning the areas of aerospace and industrial control. Such communication relies on time-triggered flows, with each flow consisting of periodic packets originating from a source and destined for a destination node. Each packet needs to reach its destination before its deadline. Different flows can have different cycle lengths. To achieve assured transmission of time-triggered flows, existing efforts constrain the packets of a flow to be cyclically transmitted along the same path. Under such Fixed Cyclic Scheduling (FCS), reservation for flows with different cycle lengths can become incompatible over a shared link, limiting the total number of admissible flows. Considering the cycle lengths of different flows, a hyper-cycle has length equal to their least common multiple (LCM). It determines the time duration over which the scheduling compatibility of the different flows can be checked. In this work, we propose a more flexible schedule scheme called the Hypercycle-level Flexible Schedule (HFS) scheme, where a flow's resource reservation can change across cycles within a hypercycle. HFS can significantly increase the number of admitted flows by providing more scheduling options while remaining perfectly compatible with existing Time-Triggered Ethernet system. We show that, theoretically the possible capacity gain provided by HFS over FCS can be unbounded. We formulate the joint pathfinding and scheduling problem under HFS as an ILP problem which we prove to be NP-Hard. To solve HFS efficiently, we further propose a least-load-first heuristic (HFS-LLF), solving HFS as a sequence of shortest path problems. Extensive study shows that HFS admits up to 6 times the number of flows achieved by FCS. Moreover, our proposed HFS-LLF can run 104 times faster than solving HFS using a generic solver.

cs.NI↗

arXiv2025

DeepMultiConnectome: Deep Multi-Task Prediction of Structural Connectomes Directly from Diffusion MRI Tractography

Diffusion MRI (dMRI) tractography enables in vivo mapping of brain structural connections, but traditional connectome generation is time-consuming and requires gray matter parcellation, posing challenges for large-scale studies. We introduce DeepMultiConnectome, a deep-learning model that predicts structural connectomes directly from tractography, bypassing the need for gray matter parcellation while supporting multiple parcellation schemes. Using a point-cloud-based neural network with multi-task learning, the model classifies streamlines according to their connected regions across two parcellation schemes, sharing a learned representation. We train and validate DeepMultiConnectome on tractography from the Human Connectome Project Young Adult dataset ($n = 1000$), labeled with an 84 and 164 region gray matter parcellation scheme. DeepMultiConnectome predicts multiple structural connectomes from a whole-brain tractogram containing 3 million streamlines in approximately 40 seconds. DeepMultiConnectome is evaluated by comparing predicted connectomes with traditional connectomes generated using the conventional method of labeling streamlines using a gray matter parcellation. The predicted connectomes are highly correlated with traditionally generated connectomes ($r = 0.992$ for an 84-region scheme; $r = 0.986$ for a 164-region scheme) and largely preserve network properties. A test-retest analysis of DeepMultiConnectome demonstrates reproducibility comparable to traditionally generated connectomes. The predicted connectomes perform similarly to traditionally generated connectomes in predicting age and cognitive function. Overall, DeepMultiConnectome provides a scalable, fast model for generating subject-specific connectomes across multiple parcellation schemes.

eess.IV↗

arXiv2025

EnerBridge-DPO: Energy-Guided Protein Inverse Folding with Markov Bridges and Direct Preference Optimization

Designing protein sequences with optimal energetic stability is a key challenge in protein inverse folding, as current deep learning methods are primarily trained by maximizing sequence recovery rates, often neglecting the energy of the generated sequences. This work aims to overcome this limitation by developing a model that directly generates low-energy, stable protein sequences. We propose EnerBridge-DPO, a novel inverse folding framework focused on generating low-energy, high-stability protein sequences. Our core innovation lies in: First, integrating Markov Bridges with Direct Preference Optimization (DPO), where energy-based preferences are used to fine-tune the Markov Bridge model. The Markov Bridge initiates optimization from an information-rich prior sequence, providing DPO with a pool of structurally plausible sequence candidates. Second, an explicit energy constraint loss is introduced, which enhances the energy-driven nature of DPO based on prior sequences, enabling the model to effectively learn energy representations from a wealth of prior knowledge and directly predict sequence energy values, thereby capturing quantitative features of the energy landscape. Our evaluations demonstrate that EnerBridge-DPO can design protein complex sequences with lower energy while maintaining sequence recovery rates comparable to state-of-the-art models, and accurately predicts $ΔΔG$ values between various sequences.

cs.LG↗

arXiv2025

A Comprehensive Survey in LLM(-Agent) Full Stack Safety: Data, Training and Deployment

The remarkable success of Large Language Models (LLMs) has illuminated a promising pathway toward achieving Artificial General Intelligence for both academic and industrial communities, owing to their unprecedented performance across various applications. As LLMs continue to gain prominence in both research and commercial domains, their security and safety implications have become a growing concern, not only for researchers and corporations but also for every nation. Currently, existing surveys on LLM safety primarily focus on specific stages of the LLM lifecycle, e.g., deployment phase or fine-tuning phase, lacking a comprehensive understanding of the entire "lifechain" of LLMs. To address this gap, this paper introduces, for the first time, the concept of "full-stack" safety to systematically consider safety issues throughout the entire process of LLM training, deployment, and eventual commercialization. Compared to the off-the-shelf LLM safety surveys, our work demonstrates several distinctive advantages: (I) Comprehensive Perspective. We define the complete LLM lifecycle as encompassing data preparation, pre-training, post-training, deployment and final commercialization. To our knowledge, this represents the first safety survey to encompass the entire lifecycle of LLMs. (II) Extensive Literature Support. Our research is grounded in an exhaustive review of over 800+ papers, ensuring comprehensive coverage and systematic organization of security issues within a more holistic understanding. (III) Unique Insights. Through systematic literature analysis, we have developed reliable roadmaps and perspectives for each chapter. Our work identifies promising research directions, including safety in data generation, alignment techniques, model editing, and LLM-based agent systems. These insights provide valuable guidance for researchers pursuing future work in this field.

cs.CR↗

arXiv2025

ViVo: A Dataset for Volumetric Video Reconstruction and Compression

As research on neural volumetric video reconstruction and compression flourishes, there is a need for diverse and realistic datasets, which can be used to develop and validate reconstruction and compression models. However, existing volumetric video datasets lack diverse content in terms of both semantic and low-level features that are commonly present in real-world production pipelines. In this context, we propose a new dataset, ViVo, for VolumetrIc VideO reconstruction and compression. The dataset is faithful to real-world volumetric video production and is the first dataset to extend the definition of diversity to include both human-centric characteristics (skin, hair, etc.) and dynamic visual phenomena (transparent, reflective, liquid, etc.). Each video sequence in this database contains raw data including fourteen multi-view RGB and depth video pairs, synchronized at 30FPS with per-frame calibration and audio data, and their associated 2-D foreground masks and 3-D point clouds. To demonstrate the use of this database, we have benchmarked three state-of-the-art (SotA) 3-D reconstruction methods and two volumetric video compression algorithms. The obtained results evidence the challenging nature of the proposed dataset and the limitations of existing datasets for both volumetric video reconstruction and compression tasks, highlighting the need to develop more effective algorithms for these applications. The database and the associated results are available at https://vivo-bvicr.github.io/

cs.CV↗

arXiv2025

Lithography defined semiconductor moires with anomalous in-gap quantum Hall states

Quantum materials and phenomena have attracted great interest for their potential applications in next-generation microelectronics and quantum-information technologies. In one especially interesting class of quantum materials, moire superlattices (MSL) formed by twisted bilayers of 2D materials, a wide range of novel phenomena are observed. However, there exist daunting challenges such as reproducibility and scalability of utilizing 2D MSLs for microelectronics and quantum technologies due to their exfoliate-tear-stack method. Here, we propose lithography defined semiconductor moires superlattices, in which three fundamental parameters, electron-electron interaction, spin-orbit coupling, and band topology, are designable. We experimentally investigate quantum transport properties in a moire specimen made in an InAs quantum well. Strong anomalous in-gap states are observed within the same integer quantum Hall state. Our work opens up new horizons for studying 2D quantum-materials phenomena in semiconductors featuring superior industry-level quality and state-of-the-art technologies, and they may potentially enable new quantum information and microelectronics technologies.

cond-mat.mes-hall↗

arXiv2025

EGA-V1: Unifying Online Advertising with End-to-End Learning

Modern industrial advertising systems commonly employ Multi-stage Cascading Architectures (MCA) to balance computational efficiency with ranking accuracy. However, this approach presents two fundamental challenges: (1) performance inconsistencies arising from divergent optimization targets and capability differences between stages, and (2) failure to account for advertisement externalities - the complex interactions between candidate ads during ranking. These limitations ultimately compromise system effectiveness and reduce platform profitability. In this paper, we present EGA-V1, an end-to-end generative architecture that unifies online advertising ranking as one model. EGA-V1 replaces cascaded stages with a single model to directly generate optimal ad sequences from the full candidate ad corpus in location-based services (LBS). The primary challenges associated with this approach stem from high costs of feature processing and computational bottlenecks in modeling externalities of large-scale candidate pools. To address these challenges, EGA-V1 introduces an algorithm and engine co-designed hybrid feature service to decouple user and ad feature processing, reducing latency while preserving expressiveness. To efficiently extract intra- and cross-sequence mutual information, we propose RecFormer with an innovative cluster-attention mechanism as its core architectural component. Furthermore, we propose a bi-stage training strategy that integrates pre-training with reinforcement learning-based post-training to meet sophisticated platform and advertising objectives. Extensive offline evaluations on public benchmarks and large-scale online A/B testing on industrial advertising platform have demonstrated the superior performance of EGA-V1 over state-of-the-art MCAs.

cs.IR↗