arXiv subjects

Xin Zhang

Publications and source records attributed to Xin Zhang.

At least 343 records · Page 19Linked to original sources

arXiv2025

Condition Weaving Meets Expert Modulation: Towards Universal and Controllable Image Generation

The image-to-image generation task aims to produce controllable images by leveraging conditional inputs and prompt instructions. However, existing methods often train separate control branches for each type of condition, leading to redundant model structures and inefficient use of computational resources. To address this, we propose a Unified image-to-image Generation (UniGen) framework that supports diverse conditional inputs while enhancing generation efficiency and expressiveness. Specifically, to tackle the widely existing parameter redundancy and computational inefficiency in controllable conditional generation architectures, we propose the Condition Modulated Expert (CoMoE) module. This module aggregates semantically similar patch features and assigns them to dedicated expert modules for visual representation and conditional modeling. By enabling independent modeling of foreground features under different conditions, CoMoE effectively mitigates feature entanglement and redundant computation in multi-condition scenarios. Furthermore, to bridge the information gap between the backbone and control branches, we propose WeaveNet, a dynamic, snake-like connection mechanism that enables effective interaction between global text-level control from the backbone and fine-grained control from conditional branches. Extensive experiments on the Subjects-200K and MultiGen-20M datasets across various conditional image generation tasks demonstrate that our method consistently achieves state-of-the-art performance, validating its advantages in both versatility and effectiveness. The code has been uploaded to https://github.com/gavin-gqzhang/UniGen.

cs.CV↗

arXiv2025

Prospects for joint multiband detection of intermediate-mass black holes by LGWA and the Einstein Telescope

Gravitational-wave (GW) detection offers a novel approach to exploring intermediate-mass black holes (IMBHs). The GW signals from IMBH mergers mainly fall in the decihertz frequency band. The lunar-based GW detector, the Lunar Gravitational-Wave Antenna (LGWA), exhibits high sensitivity in this band, making it particularly well-suited for detecting IMBHs. However, for lower-mass IMBHs, the late inspiral and merger signals enter the sensitive frequency range of ground-based GW detectors. In this work, we aim to explore how multi-band observations with LGWA and the third-generation ground-based GW detector, the Einstein Telescope (ET), can contribute to detecting the population of IMBHs. We consider three population distribution cases of IMBHs, including two population models based on astrophysical motivations and a uniform distribution, and compute the signal-to-noise ratios for LGWA, ET, and their combination to directly compare their capabilities in detecting IMBH mergers. Our results suggest that LGWA possesses strong detection capability for high-mass IMBH mergers. At redshift $z = 1$, LGWA's detection rate for IMBH binaries with primary masses above $5 \times 10^4~M_\odot$ is largely insensitive to orbital inclination and mass ratio. In contrast, ET is more suited for detecting IMBH binaries with primary masses below $10^3~M_\odot$. The multi-band observation of LGWA and ET possesses strong detection capabilities across the full IMBH mass spectrum. Furthermore, we find that the multi-band detection can significantly and effectively recover the IMBH population distributions. In summary, we conclude that the multi-band observations of LGWA and ET will provide powerful detection capabilities for IMBHs and are expected to significantly enhance our understanding of this important yet still poorly observed class of black holes.

gr-qc↗

arXiv2025

Cosmic sign-reversal: non-parametric reconstruction of interacting dark energy with DESI DR2

A direct interaction between dark energy and dark matter provides a natural and important extension to the standard $Λ$CDM cosmology. We perform a non-parametric reconstruction of the vacuum energy ($w=-1$) interacting with cold dark matter using the cosmological data from DESI DR2, Planck CMB, and three SNIa samples (PP, DESY5, and Union3). By discretizing the coupling function $β(z)$ into 20 redshift bins and assuming a Gaussian smoothness prior, we reconstruct $β(z)$ without assuming any specific parameterization. The mean reconstructed $β(z)$ changes sign during cosmic evolution, indicating an energy transfer from cold dark matter to dark energy at early times and a reverse flow at late times. At high redshifts, $β(z)$ shows a $\sim 2σ$ deviation from $Λ$CDM. At low redshifts, the results depend on the SNIa sample: CMB+DESI and CMB+DESI+PP yield $β(z)$ consistent with zero within $2σ$, while CMB+DESI+DESY5 and CMB+DESI+Union3 prefer negative $β$ at $\sim2σ$. Both $χ^2$ tests and Bayesian analyses favor the $β(z)$ model, with CMB+DESI DR2+DESY5 showing the most significant support through the largest improvement in goodness of fit ($Δχ^2_{\rm MAP}=-17.76$) and strongest Bayesian evidence ($\ln\mathcal{B} = 5.98 \pm 0.69$). Principal component analysis reveals that the data effectively constrain three additional degrees of freedom in the $β(z)$ model, accounting for most of the improvement in goodness of fit. Our results demonstrate that the dynamical dark energy preference in current data can be equally well explained by such a sign-reversal interacting dark energy, highlighting the need for future observations to break this degeneracy.

astro-ph.CO↗

arXiv2025

DISTA-Net: Dynamic Closely-Spaced Infrared Small Target Unmixing

Resolving closely-spaced small targets in dense clusters presents a significant challenge in infrared imaging, as the overlapping signals hinder precise determination of their quantity, sub-pixel positions, and radiation intensities. While deep learning has advanced the field of infrared small target detection, its application to closely-spaced infrared small targets has not yet been explored. This gap exists primarily due to the complexity of separating superimposed characteristics and the lack of an open-source infrastructure. In this work, we propose the Dynamic Iterative Shrinkage Thresholding Network (DISTA-Net), which reconceptualizes traditional sparse reconstruction within a dynamic framework. DISTA-Net adaptively generates convolution weights and thresholding parameters to tailor the reconstruction process in real time. To the best of our knowledge, DISTA-Net is the first deep learning model designed specifically for the unmixing of closely-spaced infrared small targets, achieving superior sub-pixel detection accuracy. Moreover, we have established the first open-source ecosystem to foster further research in this field. This ecosystem comprises three key components: (1) CSIST-100K, a publicly available benchmark dataset; (2) CSO-mAP, a custom evaluation metric for sub-pixel detection; and (3) GrokCSO, an open-source toolkit featuring DISTA-Net and other models. Our code and dataset are available at https://github.com/GrokCV/GrokCSO.

cs.CV↗

arXiv2025

RelServe: Fast LLM Inference Serving on Relational Data

The use of Large Language Models (LLMs) for querying relational data has given rise to relQuery, a workload pattern that applies templated LLM calls to structured tables. As relQuery services become more widely adopted in applications such as AI-powered spreadsheets, fast response times under concurrent query loads are increasingly important. Unfortunately, current LLM engines face severe latency bottlenecks from Head-of-Line (HoL) blocking across three comparable inference phases: waiting, core running, and tail running. Existing static priority scheduling methods only address HoL blocking during the waiting phase, leaving two critical problems unsolved. First, the absence of a priority update mechanism causes inaccurate prioritization and continued HoL blocking during core execution. Second, suboptimal prefill-decode batching exacerbates HoL blocking in tail execution and worsens latency trade-offs between running and waiting relQueries. To address these problems, we propose RelServe, an optimized LLM engine for low-latency relQuery serving. RelServe features two core innovations: a Dynamic Priority Updater that continuously adjusts priorities while minimizing overhead via statistical approximations, and an Adaptive Batch Arranger that quantitatively evaluates candidate prefill and decode batches to minimize projected average latency. Extensive experiments on four real-world datasets using LLMs ranging from 13B to 70B parameters show that RelServe reduces average serving latency by up to 3.1x compared to vLLM.

cs.DB↗

arXiv2025

ERF-BA-TFD+: A Multimodal Model for Audio-Visual Deepfake Detection

Deepfake detection is a critical task in identifying manipulated multimedia content. In real-world scenarios, deepfake content can manifest across multiple modalities, including audio and video. To address this challenge, we present ERF-BA-TFD+, a novel multimodal deepfake detection model that combines enhanced receptive field (ERF) and audio-visual fusion. Our model processes both audio and video features simultaneously, leveraging their complementary information to improve detection accuracy and robustness. The key innovation of ERF-BA-TFD+ lies in its ability to model long-range dependencies within the audio-visual input, allowing it to better capture subtle discrepancies between real and fake content. In our experiments, we evaluate ERF-BA-TFD+ on the DDL-AV dataset, which consists of both segmented and full-length video clips. Unlike previous benchmarks, which focused primarily on isolated segments, the DDL-AV dataset allows us to assess the model's performance in a more comprehensive and realistic setting. Our method achieves state-of-the-art results on this dataset, outperforming existing techniques in terms of both accuracy and processing speed. The ERF-BA-TFD+ model demonstrated its effectiveness in the "Workshop on Deepfake Detection, Localization, and Interpretability," Track 2: Audio-Visual Detection and Localization (DDL-AV), and won first place in this competition.

cs.AI↗

arXiv2025

Cross-Field Interface-Aware Neural Operators for Multiphase Flow Simulation

Multiphase flow simulation is critical in science and engineering but incurs high computational costs due to complex field discontinuities and the need for high-resolution numerical meshes. While Neural Operators (NOs) offer an efficient alternative for solving Partial Differential Equations (PDEs), they struggle with two core challenges unique to multiphase systems: spectral bias caused by spatial heterogeneity at phase interfaces, and the persistent scarcity of expensive, high-resolution field data. This work introduces the Interface Information Aware Neural Operator (IANO), a novel architecture that mitigates these issues by leveraging readily obtainable interface data (e.g., topology and position). Interface data inherently contains the high-frequency features not only necessary to complement the physical field data, but also help with spectral bias. IANO incorporates an interface-aware function encoding mechanism to capture dynamic coupling, and a geometry-aware positional encoding method to enhance spatial fidelity for pointwise super-resolution. Empirical results across multiple multiphase flow cases demonstrate that IANO achieves significant accuracy improvements (up to $\sim$10\%) over existing NO baselines. Furthermore, IANO exhibits superior generalization capabilities in low-data and noisy settings, confirming its utility for practical, data-efficient $\text{AI}$-based multiphase flow simulations.

physics.flu-dyn↗

arXiv2025

RoboScape-R: Unified Reward-Observation World Models for Generalizable Robotics Training via RL

Achieving generalizable embodied policies remains a key challenge. Traditional policy learning paradigms, including both Imitation Learning (IL) and Reinforcement Learning (RL), struggle to cultivate generalizability across diverse scenarios. While IL policies often overfit to specific expert trajectories, RL suffers from the inherent lack of a unified and general reward signal necessary for effective multi-scene generalization. We posit that the world model is uniquely capable of serving as a universal environment proxy to address this limitation. However, current world models primarily focus on their ability to predict observations and still rely on task-specific, handcrafted reward functions, thereby failing to provide a truly general training environment. Toward this problem, we propose RoboScape-R, a framework leveraging the world model to serve as a versatile, general-purpose proxy for the embodied environment within the RL paradigm. We introduce a novel world model-based general reward mechanism that generates ''endogenous'' rewards derived from the model's intrinsic understanding of real-world state transition dynamics. Extensive experiments demonstrate that RoboScape-R effectively addresses the limitations of traditional RL methods by providing an efficient and general training environment that substantially enhances the generalization capability of embodied policies. Our approach offers critical insights into utilizing the world model as an online training strategy and achieves an average 37.5% performance improvement over baselines under out-of-domain scenarios.

cs.RO↗

arXiv2025

A $1000\times$ Faster LLM-enhanced Algorithm For Path Planning in Large-scale Grid Maps

Path planning in grid maps, arising from various applications, has garnered significant attention. Existing methods, such as A*, Dijkstra, and their variants, work well for small-scale maps but fail to address large-scale ones due to high search time and memory consumption. Recently, Large Language Models (LLMs) have shown remarkable performance in path planning but still suffer from spatial illusion and poor planning performance. Among all the works, LLM-A* \cite{meng2024llm} leverages LLM to generate a series of waypoints and then uses A* to plan the paths between the neighboring waypoints. In this way, the complete path is constructed. However, LLM-A* still suffers from high computational time for large-scale maps. To fill this gap, we conducted a deep investigation into LLM-A* and found its bottleneck, resulting in limited performance. Accordingly, we design an innovative LLM-enhanced algorithm, abbr. as iLLM-A*. iLLM-A* includes 3 carefully designed mechanisms, including the optimization of A*, an incremental learning method for LLM to generate high-quality waypoints, and the selection of the appropriate waypoints for A* for path planning. Finally, a comprehensive evaluation on various grid maps shows that, compared with LLM-A*, iLLM-A* \textbf{1) achieves more than $1000\times$ speedup on average, and up to $2349.5\times$ speedup in the extreme case, 2) saves up to $58.6\%$ of the memory cost, 3) achieves both obviously shorter path length and lower path length standard deviation.}

cs.RO↗

arXiv2025

Evaluating Large Language Models for Radiology Natural Language Processing

The rise of large language models (LLMs) has marked a pivotal shift in the field of natural language processing (NLP). LLMs have revolutionized a multitude of domains, and they have made a significant impact in the medical field. Large language models are now more abundant than ever, and many of these models exhibit bilingual capabilities, proficient in both English and Chinese. However, a comprehensive evaluation of these models remains to be conducted. This lack of assessment is especially apparent within the context of radiology NLP. This study seeks to bridge this gap by critically evaluating thirty two LLMs in interpreting radiology reports, a crucial component of radiology NLP. Specifically, the ability to derive impressions from radiologic findings is assessed. The outcomes of this evaluation provide key insights into the performance, strengths, and weaknesses of these LLMs, informing their practical applications within the medical domain.

cs.CL↗

arXiv2025

On The Role of Pretrained Language Models in General-Purpose Text Embeddings: A Survey

Text embeddings have attracted growing interest due to their effectiveness across a wide range of natural language processing (NLP) tasks, including retrieval, classification, clustering, bitext mining, and summarization. With the emergence of pretrained language models (PLMs), general-purpose text embeddings (GPTE) have gained significant traction for their ability to produce rich, transferable representations. The general architecture of GPTE typically leverages PLMs to derive dense text representations, which are then optimized through contrastive learning on large-scale pairwise datasets. In this survey, we provide a comprehensive overview of GPTE in the era of PLMs, focusing on the roles PLMs play in driving its development. We first examine the fundamental architecture and describe the basic roles of PLMs in GPTE, i.e., embedding extraction, expressivity enhancement, training strategies, learning objectives, and data construction. We then describe advanced roles enabled by PLMs, including multilingual support, multimodal integration, code understanding, and scenario-specific adaptation. Finally, we highlight potential future research directions that move beyond traditional improvement goals, including ranking integration, safety considerations, bias mitigation, structural information incorporation, and the cognitive extension of embeddings. This survey aims to serve as a valuable reference for both newcomers and established researchers seeking to understand the current state and future potential of GPTE.

cs.CL↗

arXiv2025

AnchorOPT: Towards Optimizing Dynamic Anchors for Adaptive Prompt Learning

Existing prompt learning methods, which are built upon CLIP models, leverage textual tokens as anchors to guide the learnable soft tokens. This guidance improves CLIP generalizations. However, these anchors-static in both value and position-lack cross-task and stage-adaptive flexibility. To address this limitation, we propose AnchorOPT, a dynamic anchor-based prompt learning framework. Specifically, AnchorOPT introduces dynamism in two key dimensions: (i) anchor values eschew handcrafted explicit textual tokens (e.g., "shape", "color"), instead learning dynamically from task-specific data; and (ii) the positional relationship between anchor and soft tokens is no longer fixed but adaptively optimized via a learnable position matrix conditioned on the training stage and task context. Training occurs in two stages: we first learn the anchor tokens, then freeze and transfer them to the second stage for optimization of soft tokens and the position matrix. Extensive experiments demonstrate that using only a simple learnable anchor and position matrix achieves performance comparable to or exceeding some methods incorporating additional learnable modules or regularization techniques. As a plug-and-play module, AnchorOPT integrates seamlessly into existing frameworks, yielding consistent performance gains across diverse datasets. Code is publicly available at https://github.com/zhengli97/ATPrompt.

cs.CV↗

arXiv2025

Testing the cosmic distance duality relation with baryon acoustic oscillations and supernovae data

One of the most fundamental relationships in modern cosmology is the cosmic distance duality relation (CDDR), which describes the relationship between the angular diameter distance ($D_{\rm A}$) and the luminosity distance ($D_{\rm L}$), and is expressed as: $η(z)=D_{\rm L}(z)(1+z)^{-2}/D_{\rm A}(z)=1$. In this work, we conduct a comprehensive test of the CDDR by combining baryon acoustic oscillation (BAO) data from the SDSS and DESI surveys with type Ia supernova (SN) data from PantheonPlus and DESY5. We utilize an artificial neural network approach to match the SN and BAO data at the same redshift. To explore potential violations of the CDDR, we consider three different parameterizations: (i) $η(z)=1+η_0z$; (ii) $η(z)=1+η_0z/(1+z)$; (iii) $η(z)=1+η_0\ln(1+z)$. Our results indicate that the calibration of the SN absolute magnitude $M_{\rm B}$ plays a crucial role in testing potential deviations from the CDDR, as there exists a significant negative correlation between $η_0$ and $M_{\rm B}$. For PantheonPlus analysis, when $M_{\rm B}$ is treated as a free parameter, no evidence of CDDR violation is found. In contrast, fixing $M_{\rm B}$ to the $M_{\rm B}^{\rm D20}$ prior with $-19.230\pm0.040$ mag leads to a deviation at approximately the $2σ$ level, while fixing $M_{\rm B}$ to the $M_{\rm B}^{\rm B23}$ prior with $-19.396\pm0.016$ mag remains in agreement with the CDDR. Furthermore, overall analyses based on the SDSS+DESY5 and DESI+DESY5 data consistently show no evidence of the deviation from the CDDR.

astro-ph.CO↗

arXiv2025

Prospects for cosmological research using hundred-meter-class radio telescopes: 21-cm intensity mapping survey strategies with QTT, JRT, and HRT

Understanding dark energy requires precision measurements of the expansion history of the universe and the growth of large-scale structure. The 21 cm intensity mapping (21 cm IM) technique enables rapid large-area surveys that can deliver these measurements. China is constructing three hundred-meter-class single-dish radio telescopes, including the QiTai 110 m Radio Telescope (QTT), the 120 m Jingdong Radio Telescope (JRT), and the 120 m Huadian Radio Telescope (HRT), whose designs are well suited for 21 cm IM cosmology. We use a Fisher-to-MCMC forecasting framework to evaluate the baryon acoustic oscillations / redshift space distortions (BAO/RSD) measurement capabilities of QTT, JRT, and HRT and propagate them to dark-energy constraints in the $w_0w_a$CDM model. Our results show that achieving a redshift coverage up to $z_{\mathrm{max}} = 1$ is crucial for fully realising the potential of hundred-meter-class single-dish telescopes for 21 cm cosmology. If all three telescopes carry out 21 cm IM surveys over the same redshift range up to $z_{\mathrm{max}}=1$ and combine their BAO/RSD measurements, QTT+JRT+HRT yield $σ(w_0)=0.094$ and $σ(w_a)=0.487$, providing tighter constraints than DESI DR2 results.

astro-ph.CO↗

arXiv2025

Comparison for semi-continuous viscosity solutions for second order PDEs on the Wasserstein space

In this paper, we prove a comparison result for semi-continuous viscosity solutions of a class of second-order PDEs in the Wasserstein space. This allows us to remove the Lipschitz continuity assumption with respect to the Fourier-Wasserstein distance in AriX: 2309.05040 and obtain uniqueness by directly working in the Wasserstein space. In terms of its application, we characterize the value function of a stochastic control problem with partial observation as the unique viscosity solution to its corresponding HJB equation. Additionally, we present an application to a prediction problem under partial monitoring, where we establish an upper bound on the limit of regret using our comparison principle for degenerate dynamics.

math.AP↗

arXiv2025

Covariate Adjustment in Randomized Experiments Motivated by Higher-Order Influence Functions

Higher-Order Influence Functions (HOIF), developed in a series of papers over the past twenty years, are a fundamental theoretical device for constructing rate-optimal causal-effect estimators from observational studies. However, the value of HOIF for analyzing well-conducted randomized controlled trials (RCT) has not been explicitly explored. In the recent U.S. Food and Drug Administration and European Medicines Agency guidelines on the practice of covariate adjustment in analyzing RCT, in addition to the simple, unadjusted difference-in-mean estimator, it was also recommended to report the estimator adjusting for baseline covariates via a simple parametric working model, such as a linear model. However, when the number of baseline covariates $p$ is large, the recommendation is somewhat murky. In this paper, we show that HOIF-motivated estimators for the treatment-specific mean have significantly improved statistical properties compared to popular adjusted estimators in practice when $p$ is relatively large relative to the sample size $n$. We also characterize the conditions under which the HOIF-motivated estimator improves upon the unadjusted one. More importantly, we demonstrate that several state-of-the-art adjusted estimators proposed recently can be interpreted as particular HOIF-motivated estimators, thereby placing these estimators in a more unified framework. Numerical and empirical studies are conducted to corroborate our theoretical findings. An accompanying R package can be found on CRAN.

stat.ME↗

arXiv2025

Mock Observations for the CSST Mission: Main Surveys-the Slitless Spectroscopy Simulation

The China Space Station Telescope (CSST), slated to become China's largest space-based optical telescope in the coming decade, is designed to conduct wide-field sky surveys with high spatial resolution. Among its key observational modes, slitless spectral observation allows simultaneous imaging and spectral data acquisition over a wide field of view, offering significant advantages for astrophysical studies. Currently, the CSST is in the development phase and lacks real observational data. As a result, the development of its data processing pipeline and scientific pre-research must rely on the mock data generated through simulations. This work focuses on developing a simulation framework for the CSST slitless spectral imaging system, analyzing its spectral dispersing properties and structural design. Additionally, the detection performance of the slitless spectral system is assessed for various astrophysical targets. Simulation results demonstrate that nearly all 1st order spectra are accompanied by corresponding 0th order images, facilitating accurate source identification. Furthermore, the GI spectral band exhibits superior detection efficiency compared to the GV and GU bands, establishing it as the primary observational band for stellar and galactic studies. This work successfully develops a simulation framework for the CSST slitless spectroscopic equipment.

astro-ph.IM↗

arXiv2025

Secrecy Analysis for MISO Broadcast Systems with Regularized Zero-Forcing Precoding

As an effective way to enhance the physical layer security (PLS) for the broadcast channel (BC), regularized zero-forcing (RZF) precoding has attracted much attention. However, the reliability performance, i.e., secrecy outage probability (SOP), of RZF is not well investigated in the literature. In this paper, we characterize the secrecy performance of RZF precoding in the large multiple-input single-output (MISO) broadcast system. For this purpose, we first consider a central limit theorem (CLT) for the joint distribution of the users' signal-to-interference-plus-noise ratio (SINR) and the eavesdropper's (Eve's) signal-to-noise ratio (ESNR) by leveraging random matrix theory (RMT). The result is then utilized to obtain a closed-form approximation for the ergodic secrecy rate (ESR) and SOP of three typical scenarios: the case with only external Eves, the case with only internal Eves, and that with both. The derived results are then used to evaluate the percentage of users in secrecy outage and the required number of transmit antennas to achieve a positive secrecy rate. It is shown that, with equally-capable Eves, the secrecy loss caused by external Eves is higher than that caused by internal Eves. Numerical simulations validate the accuracy of the theoretical results.

cs.IT↗