arXiv subjects

Xiao Zhang

Publications and source records attributed to Xiao Zhang.

At least 145 records · Page 8Linked to original sources

arXiv2025

Unlocking Multi-View Insights in Knowledge-Dense Retrieval-Augmented Generation

While Retrieval-Augmented Generation (RAG) plays a crucial role in the application of Large Language Models (LLMs), existing retrieval methods in knowledge-dense domains like law and medicine still suffer from a lack of multi-perspective views, which are essential for improving interpretability and reliability. Previous research on multi-view retrieval often focused solely on different semantic forms of queries, neglecting the expression of specific domain knowledge perspectives. This paper introduces a novel multi-view RAG framework, MVRAG, tailored for knowledge-dense domains that utilizes intention-aware query rewriting from multiple domain viewpoints to enhance retrieval precision, thereby improving the effectiveness of the final inference. Experiments conducted on legal and medical case retrieval demonstrate significant improvements in recall and precision rates with our framework. Our multi-perspective retrieval approach unleashes the potential of multi-view information enhancing RAG tasks, accelerating the further application of LLMs in knowledge-intensive fields.

cs.CL↗

arXiv2025

CAPO: Towards Enhancing LLM Reasoning through Generative Credit Assignment

Reinforcement Learning with Verifiable Rewards (RLVR) has improved the reasoning abilities of Large Language Models (LLMs) by using rule-based binary feedback. However, current RLVR methods typically assign the same reward to every token. This coarse-grained feedback hampers precise credit assignment, making it hard for models to identify which reasoning steps lead to success or failure, and often results in suboptimal policies. Methods like PPO provide credit assignment by value estimation, but yield inaccurate and unverifiable signals due to limited sampling. On the other hand, methods using Process Reward Models can provide step-wise rewards but suffer from several key limitations: they require high-quality process supervision labels, the feedback is unreliable due to probabilistic reward modeling, and their application in online reinforcement learning (RL) is time-consuming. To overcome these limitations, we introduce a simple but efficient method-Credit Assignment Policy Optimization (CAPO). Instead of training auxiliary models, CAPO directly leverages an off-the-shelf, general-purpose LLM as a Generative Process Reward Model (LLM-as-GenPRM) to generate all step-wise critique by one pass only based on the correctness of the step itself, providing deterministic token-level credits to refine the tokens that were originally assigned identical rule-based rewards. To further enhance the accuracy and robustness, we employ voting mechanisms that scale with the number of generated critiques. Extensive experiments on various backbones like Llama and Qwen models show that CAPO consistently outperforms supervised learning-based and RL-based fine-tuning methods across four challenging mathematical benchmarks and three out-of-domain benchmarks. Further analysis shows that CAPO can help the model to foster the learning of correct reasoning pathways leading to correct answers.

cs.LG↗

arXiv2025

HyperDiffusionFields (HyDiF): Diffusion-Guided Hypernetworks for Learning Implicit Molecular Neural Fields

We introduce HyperDiffusionFields (HyDiF), a framework that models 3D molecular conformers as continuous fields rather than discrete atomic coordinates or graphs. At the core of our approach is the Molecular Directional Field (MDF), a vector field that maps any point in space to the direction of the nearest atom of a particular type. We represent MDFs using molecule-specific neural implicit fields, which we call Molecular Neural Fields (MNFs). To enable learning across molecules and facilitate generalization, we adopt an approach where a shared hypernetwork, conditioned on a molecule, generates the weights of the given molecule's MNF. To endow the model with generative capabilities, we train the hypernetwork as a denoising diffusion model, enabling sampling in the function space of molecular fields. Our design naturally extends to a masked diffusion mechanism to support structure-conditioned generation tasks, such as molecular inpainting, by selectively noising regions of the field. Beyond generation, the localized and continuous nature of MDFs enables spatially fine-grained feature extraction for molecular property prediction, something not easily achievable with graph or point cloud based methods. Furthermore, we demonstrate that our approach scales to larger biomolecules, illustrating a promising direction for field-based molecular modeling.

cs.LG↗

arXiv2025

Merge and Guide: Unifying Model Merging and Guided Decoding for Controllable Multi-Objective Generation

Adapting to diverse user needs at test time is a key challenge in controllable multi-objective generation. Existing methods are insufficient: merging-based approaches provide indirect, suboptimal control at the parameter level, often disregarding the impacts of multiple objectives. While decoding-based guidance is more direct, it typically requires aggregating logits from multiple expert models, incurring significant space overhead and relying heavily on individual model capacity. To address these issues, we introduce Merge-And-GuidE (MAGE), a two-stage framework that leverages model merging for guided decoding. We first identify a critical compatibility problem between the guidance and base models. In Stage 1, MAGE resolves this by dynamically constructing a more robust base model, merging a series of backbone models that account for multiple objectives. In Stage 2, we merge explicit and implicit value models into a unified guidance proxy, which then steers the decoding of the base model from Stage 1. Our analysis empirically validates Linear Mode Connectivity (LMC) in value models, explores the relationship between model merging and prediction ensembling, and demonstrates the enhanced controllability afforded by our approach. Extensive experiments show that our method outperforms existing approaches, achieving superior controllability, Pareto-optimal performance, and enhanced adaptability.

cs.LG↗

arXiv2025

From Gaze to Insight: Bridging Human Visual Attention and Vision Language Model Explanation for Weakly-Supervised Medical Image Segmentation

Medical image segmentation remains challenging due to the high cost of pixel-level annotations for training. In the context of weak supervision, clinician gaze data captures regions of diagnostic interest; however, its sparsity limits its use for segmentation. In contrast, vision-language models (VLMs) provide semantic context through textual descriptions but lack the explanation precision required. Recognizing that neither source alone suffices, we propose a teacher-student framework that integrates both gaze and language supervision, leveraging their complementary strengths. Our key insight is that gaze data indicates where clinicians focus during diagnosis, while VLMs explain why those regions are significant. To implement this, the teacher model first learns from gaze points enhanced by VLM-generated descriptions of lesion morphology, establishing a foundation for guiding the student model. The teacher then directs the student through three strategies: (1) Multi-scale feature alignment to fuse visual cues with textual semantics; (2) Confidence-weighted consistency constraints to focus on reliable predictions; (3) Adaptive masking to limit error propagation in uncertain areas. Experiments on the Kvasir-SEG, NCI-ISBI, and ISIC datasets show that our method achieves Dice scores of 80.78%, 80.53%, and 84.22%, respectively-improving 3-5% over gaze baselines without increasing the annotation burden. By preserving correlations among predictions, gaze data, and lesion descriptions, our framework also maintains clinical interpretability. This work illustrates how integrating human visual attention with AI-generated semantic context can effectively overcome the limitations of individual weak supervision signals, thereby advancing the development of deployable, annotation-efficient medical AI systems. Code is available at: https://github.com/jingkunchen/FGI.

cs.CV↗

arXiv2025

Invisibility Cloak: Disappearance under Human Pose Estimation via Backdoor Attacks

Despite being significant in autonomous systems, Human Pose Estimation (HPE)'s potential risks to adversarial attacks have not received comparable attention with image classification or segmentation tasks. In this paper, we study the vulnerability of HPE systems to disappearance attacks, where the attacker aims to subtly alter the HPE training process via backdoor techniques so that any input image with some specific trigger will not be recognized as involving any human pose. As humans are typically at the center of HPE systems, a successful attack will severely threaten pedestrians' lives if a self-driving car incorrectly understands the front scene. To achieve the adversarial goal of disappearance, we propose \emph{IntC}, a general framework to craft an invisibility cloak in the HPE domain. By designing target HPE labels that do not represent any human pose, we propose three specific backdoor attacks based on our IntC framework. IntC-S and IntC-E, respectively designed for regression- and heatmap-based HPE techniques, concentrate the keypoints of triggered images in a tiny, imperceptible region. Further, to improve the attack's stealthiness, IntC-L designs the target poisons to capture the label outputs of typical landscape images without a human involved, achieving disappearance and reducing detectability simultaneously. Extensive experiments demonstrate the effectiveness and generalizability of our IntC methods in achieving the disappearance goal. By revealing the vulnerability of HPE to disappearance and backdoor attacks, we hope our work can raise awareness of the potential risks when HPE models are deployed in real-world applications.

cs.CR↗

arXiv2025

OntoURL: A Benchmark for Evaluating Large Language Models on Symbolic Ontological Understanding, Reasoning and Learning

Large language models have demonstrated remarkable capabilities across a wide range of tasks, yet their ability to process structured symbolic knowledge remains underexplored. To address this gap, we propose a taxonomy of ontological capabilities and introduce OntoURL, the first comprehensive benchmark designed to systematically evaluate LLMs' capabilities in handling ontologies -- formal and symbolic representations of domain knowledge. Based on the proposed taxonomy, OntoURL systematically assesses three dimensions: understanding, reasoning, and learning through 15 distinct tasks comprising 57,303 questions derived from 40 ontologies across 8 domains. Experiments with 20 open-source LLMs reveal significant performance differences across models, tasks, and domains, with current LLMs showing capabilities in understanding ontological knowledge but weaknesses in reasoning and learning tasks. Further experiments with few-shot and chain-of-thought prompting illustrate how different prompting strategies affect model performance. Additionally, a human evaluation reveals that LLMs outperform humans in understanding and reasoning tasks but fall short in most learning tasks. These findings highlight both the potential and limitations of LLMs in processing symbolic knowledge and establish OntoURL as a critical benchmark for advancing the integration of LLMs with formal knowledge representations.

cs.CL↗

arXiv2025

Highly Efficient Room-Temperature Nonvolatile Magnetic Switching by Current in Fe3GaTe2 Thin Flakes

Effectively tuning magnetic state by using current is essential for novel spintronic devices. Magnetic van der Waals (vdW) materials have shown superior properties for the applications of magnetic information storage based on the efficient spin torque effect. However, for most of known vdW ferromagnets, the ferromagnetic transition temperatures lower than room temperature strongly impede their applications and the room-temperature vdW spintronic device with low energy consumption is still a long-sought goal. Here, we realize the highly efficient room-temperature nonvolatile magnetic switching by current in a single-material device based on vdW ferromagnet Fe3GaTe2. Moreover, the switching current density and power dissipation are about 300 and 60000 times smaller than conventional spin-orbit-torque devices of magnet/heavy-metal heterostructures. These findings make an important progress on the applications of magnetic vdW materials in the fields of spintronics and magnetic information storage.

cond-mat.mtrl-sci↗

arXiv2025

Cubic BeB$_2$: A metastable $p$-type conductive material from first principles

Boron forms a wide variety of compounds with alkaline earth elements due to its unique bonding characteristics. Among these, binary compounds of Be and B display particularly rich structural diversity, attributed to the small atomic size of Be. Cubic BeB$_2$ is a particularly interesting phase, where Be donates electrons to stabilize a diamond-like boron network under high pressure. In this work, we employ \textit{ab initio} methods to conduct a detailed investigation of cubic BeB$_2$ and its functional properties. We show that this metastable phase is dynamically stable under ambient conditions, and its lattice match to existing substrate materials suggests possible epitaxial stabilization via thin-film growth routes. Through a comprehensive characterization of its electronic, transport, and superconductivity properties, we demonstrate that cubic BeB$_2$ exhibits high hole concentrations and high hole mobility, making it a potential candidate for efficient $p$-type transport. In addition, cubic BeB$_2$ is found to exhibit low-temperature superconductivity at degenerate doping levels, similar to several other doped covalent semiconductors such as diamond, Si, and SiC.

cond-mat.mtrl-sci↗

arXiv2025

Controllable Adversarial Makeup for Privacy via Text-Guided Diffusion

As face recognition becomes more widespread in government and commercial services, its potential misuse raises serious concerns about privacy and civil rights. To counteract this threat, various anti-facial recognition techniques have been proposed, which protect privacy by adversarially perturbing face images. Among these, generative makeup-based approaches are the most widely studied. However, these methods, designed primarily to impersonate specific target identities, can only achieve weak dodging success rates while increasing the risk of targeted abuse. In addition, they often introduce global visual artifacts or a lack of adaptability to accommodate diverse makeup prompts, compromising user satisfaction. To address the above limitations, we develop MASQUE, a novel diffusion-based framework that generates localized adversarial makeups guided by user-defined text prompts. Built upon precise null-text inversion, customized cross-attention fusion with masking, and a pairwise adversarial guidance mechanism using images of the same individual, MASQUE achieves robust dodging performance without requiring any external identity. Comprehensive evaluations on open-source facial recognition models and commercial APIs demonstrate that MASQUE significantly improves dodging success rates over all baselines, along with higher perceptual fidelity preservation, stronger adaptability to various makeup prompts, and robustness to image transformations.

cs.CV↗

arXiv2025

Evolution of ultra-flat band in van der Waals kagome semiconductor Pd3P2(S1-xSex)8

We investigate the evolutions of structural parameters, optical properties, and electronic structures of van der Waals kagome semiconductor Pd3P2S8 with Se doping. When the doping level of Se increases, the bandgaps of Pd3P2(S1-xSex)8 single crystals decrease gradually, accompanying with the expanded unit cells. The first-principles calculations show that there is a flat band (FB) near the Fermi level in bulk Pd3P2S8. This FB mainly originates from the dz2-like orbitals of Pd atoms in the Pd kagome lattice, which has a finite interlayer electron hopping perpendicular to the PdS4 square plane. The interlayer hopping can be reinforced with the Se doping, inducing a stronger interlayer coupling via the chalcogen atoms at apical sites, which reduces the bandgap and enhances the cleavage energy. In contrast, the vanishing interlayer hopping in the two-dimensional limit results in the formation of ultra-FB in the monolayers of these compounds. The easy exfoliation and the existence of unique ultra-FB near EF make Pd3P2(S1-xSex)8 a model system to explore the exotic physics of FB in two-dimensional kagome lattice.

cond-mat.mtrl-sci↗

arXiv2025

ScaleCUA: Scaling Open-Source Computer Use Agents with Cross-Platform Data

Vision-Language Models (VLMs) have enabled computer use agents (CUAs) that operate GUIs autonomously, showing great potential, yet progress is limited by the lack of large-scale, open-source computer use data and foundation models. In this work, we introduce ScaleCUA, a step toward scaling open-source CUAs. It offers a large-scale dataset spanning 6 operating systems and 3 task domains, built via a closed-loop pipeline uniting automated agents with human experts. Trained on this scaled-up data, ScaleCUA can operate seamlessly across platforms. Specifically, it delivers strong gains over baselines (+26.6 on WebArena-Lite-v2, +10.7 on ScreenSpot-Pro) and sets new state-of-the-art results (94.4% on MMBench-GUI L1-Hard, 60.6% on OSWorld-G, 47.4% on WebArena-Lite-v2). These findings underscore the power of data-driven scaling for general-purpose computer use agents. We will release data, models, and code to advance future research: https://github.com/OpenGVLab/ScaleCUA.

cs.CV↗

arXiv2025

Absence of Radio Emission Reveals an Exceptionally Weak Explosion of the Putative Historical Supernova Pa 30

We present the first deep radio continuum observations of Pa 30, a nebula hosting a unique optical source driven by an ultrafast outflow with a velocity of 16,000 km s$^{-1}$. The nebula was proposed to be the remnant of a white dwarf merger that occurred in 1181CE. We report no detection of the radio diffuse emission from Pa 30 or radio emission from the central source, setting $3σ$ upper limit flux densities of $0.84\,\rm mJy$ and $0.29\,\rm mJy$ at 1.5 GHz and 6 GHz, respectively, for Pa 30. The radio surface brightness of Pa 30 is $\sim 3$ orders of magnitude smaller than that of typical supernova remnants (SNRs) with comparable angular size. If Pa 30 is an SNR, our observations show it to be the faintest known in the radio band. Considering that 10\% of the supernova (SN) kinetic energy is transferred to cosmic rays (CRs), the absence of radio synchrotron emission suggests that the SN kinetic energy $\lesssim3\times 10^{47}(B/10 μ\rm G)^{-1.65}$ erg, which is 3 to 4 orders of magnitude lower than that of typical SNRs and the lowest measured among Galactic SNRs. There is also an indication of inefficient CR acceleration for this source. The low SN kinetic energy either implies the potential existence of many more radio-faint, sub-energetic SNRs in our Galaxy or challenges the SNR interpretation of Pa 30.

astro-ph.HE↗

arXiv2025

Fermi-LAT Detected Gamma-ray Emission Likely Associated with SNR Kes 78

We have analyzed the GeV gamma-ray emission in the region of the supernova remnant (SNR) Kes 78 using $\sim$16.7 years of Fermi-LAT observations and found that the catalog sources 4FGL J1852.4+0037e and 4FGL J1851.8$-$0007c are better represented as two extended sources modeled as `2Ext'. One of them, designated as E2, is located at R.A.$=282.86^\circ$, Dec.$=-0.11^\circ$ with the 68\% containment radius $R_{68} = 0.31^\circ$, and is detected with a significance of 15.2$σ$ in the 0.2--500 GeV energy range. The gamma-ray emission of source E2 is well described by a log-parabola (LogP) spectral model with spectral index $Γ$ = 1.2 and curvature $β$ = 0.3. The fitting with electron-proton number ratio $K_{\rm ep}=0.01$ indicates that the GeV emission of source E2 is dominated by hadronic emission. Given the dense molecular environment surrounding the middle-aged SNR Kes 78, the hadronic scenario provides a natural explanation for the observed GeV emission. The extended source E2 can also be replaced with two point sources. One of them, designated as PTS1, is coincident with the newly discovered PSR J1852$-$0002g within the 68\% positional uncertainty circle, indicating a possible gamma-ray contribution from this PSR. The gamma-ray spectrum of source PTS1 can be well described by a LogP spectral shape. The synchro-curvature radiation model provides a satisfactory spectral fit for source PTS1, suggesting that some of the GeV emission from the Kes 78 region might possibly originate from the magnetosphere of PSR J1852$-$0002g.

astro-ph.HE↗

arXiv2025

GeV Gamma-Rays from Molecular Clouds Illuminated by Particles Diffusing from the Adjacent Supernova Remnant G335.2+0.1 Confined in an Expanding Bubble

We report the detection of GeV gamma-ray emission likely associated with supernova remnant (SNR) G335.2+0.1 and the finding of a molecular cloud ($\sim20$--$30^\prime$ in angular size) that is very likely in physical contact with the SNR and responsible for the gamma-ray emission. Using the 16.8 yr Fermi-LAT data, an extended emission, with a significance of 13.5 $σ$ and a radius 0.24° in 0.2--500 GeV in the uniform-disk model, was found to the adjacent east of the SNR. With archival Mopra CO-line data, a large molecular clump at local-standard-of-rest velocity $\sim-48$ to $-43$ km s$^{-1}$ was revealed appearing coincident with the gamma-ray source. The SNR was found located in a cavity encircled by a 'C'-shaped ring-like molecular shell at $-45$ to $-43$ km s$^{-1}$. This morphological agreement, together with the position-velociy diagrams made along lines cutting across the cavity, suggests that the SNR was evolving in the expanding molecular bubble created by the stellar wind of the progenitor with a mass $\gtrsim 15 M_{\mathrm{sun}}$. The giant molecular cloud, visible at around $-46$ km s$^{-1}$, and the associated SNR are thus estimated to lie at a kinematic distance of 3.1 kpc, with the HI absorption taken into account. We suggest that the SNR has entered the radiative phase after the blastwave recently struck the cavity wall. With the evolutionary scenario of the SNR, we demonstrate that the gamma-ray emission reported here can be naturally interpreted by the hadronic interaction between the accelerated protons that escaped from the SNR shock and the eastern large molecular clump.

astro-ph.HE↗

arXiv2025

Observatory Science with eXTP

Scheduled for launch in 2030, the enhanced X-ray Timing and Polarization (eXTP) telescope is a Chinese space-based mission aimed at studying extreme conditions and phenomena in astrophysics. eXTP will feature three main payloads: Spectroscopy Focusing Arrays (SFAs), Polarimetry Focusing Arrays (PFAs), and a Wide-field Camera (W2C). This white paper outlines observatory science, incorporating key scientific advances and instrumental changes since the publication of the previous white paper [1]. We will discuss perspectives of eXTP on the research domains of flare stars, supernova remnants, pulsar wind nebulae, cataclysmic variables, X-ray binaries, ultraluminous X-ray sources, AGN, and pulsar-based positioning and timekeeping.

astro-ph.IM↗

arXiv2025

A Lightweight Framework for Trigger-Guided LoRA-Based Self-Adaptation in LLMs

Large language models are unable to continuously adapt and learn from new data during reasoning at inference time. To address this limitation, we propose that complex reasoning tasks be decomposed into atomic subtasks and introduce SAGE, a trigger-guided dynamic fine-tuning framework that enables adaptive updates during reasoning at inference time. SAGE consists of three key components: (1) a Trigger module that detects reasoning failures through multiple evaluation metrics in real time; (2) a Trigger Buffer module that clusters anomaly samples using a streaming clustering process with HDBSCAN, followed by stability checks and similarity-based merging; and (3) a Lora Store module that dynamically optimizes parameter updates with an adapter pool for knowledge retention. Evaluation results show that SAGE demonstrates excellent accuracy, robustness, and stability on the atomic reasoning subtask through dynamic knowledge updating during test time.

cs.CL↗

arXiv2025

Observer-Based Realization of Control Systems

A novel model reduction framework for large-scale complex systems is proposed by introducing function-type dynamic control systems via the dimension-keeping semi-tensor product (DK-STP) of matrices. Utilizing bridge matrices, the DK-STP facilitates the construction of an approximate observer-based realization (OR) of a linear control system in the form of a function-type control system, where the functions serve as observers. A necessary and sufficient condition is established for the OR-system to admit exact observer dynamics. When an exact OR-system does not exist, an extended OR-system is developed by incorporating the original system's observers into its state. Furthermore, a minimal feedback extended OR-system is constructed, and its relationship to Kalman's minimal realization is analyzed. Finally, the proposed approach is extended to nonlinear control-affine systems.

math.OC↗