arXiv subjects

Xiang Zhang

Publications and source records attributed to Xiang Zhang.

At least 55 records · Page 3Linked to original sources

arXiv2026

Trojaning the Alignment: Stealthy Backdoor Attacks against Graph Foundation Models

Graph Foundation Models (GFMs) on text-attributed graphs (TAGs) align graph representations with language semantics to support transferable graph learning. Despite these advantages, the backdoor vulnerability of GFMs on TAGs remains insufficiently understood, especially under graph-language alignment, where graph and text representations are trained to constrain each other in a shared semantic space. Existing backdoor attacks mainly target either the graph side or the text side, treating the two modalities independently. This makes direct adaptation ineffective: graph-only triggers can be constrained by clean text semantics, while text-only triggers alter the language view but do not directly shift the graph representation being aligned and scored. TAGs also impose a stealth challenge because triggers are exposed as both node text and local graph structure, making incoherent trigger attributes or anomalous subgraphs easy to inspect or filter. In this paper, we propose STAG, a stealthy trojan attack framework designed for the graph-language alignment interface of GFMs on TAGs. STAG coordinates a graph-trigger generator with a text-side soft prompt so that trigger-attached graph representations and triggered text representations move toward the same target-class text region. To address TAG-specific stealthiness, STAG realizes trigger nodes as readable text through candidate retrieval and regularizes the trigger-attached subgraph so that its local structure remains close to the original subgraph. Extensive experiments on multiple TAG datasets and representative GFMs demonstrate the effectiveness and stealthiness of STAG. Our code is available at https://github.com/ventr1c/STAG.

cs.LG↗

arXiv2026

Multi-Granularity Context-Enhanced RAG over Multimodal Knowledge Graphs

Retrieval-augmented generation (RAG) is widely used to mitigate hallucination issues in large language models (LLMs) and multimodal large language models (MLLMs). In particular, knowledge graph (KG)-based RAG leverages structured knowledge to provide (M)LLMs with high-quality external information. Building on these works, recent studies have explored multimodal knowledge graphs (MMKGs) as knowledge bases for GraphRAG. This enables Graph RAG to integrate knowledge across multiple modalities, thereby further enhancing its performance. However, existing MMKG-based RAG methods generally follow a common pipeline in which different modalities are largely processed independently before being fusion. As a result, textual context is only used to a limited extent during visual information extraction and subsequent multimodal knowledge fusion. This brings a semantic gap between images and text which limits the multimodal GraphRAG performance. To address this issue, we propose a novel framework for constructing a Context-Enhanced MMKG (CEMMKG) to better support multimodal GraphRAG. The proposed CEMMKG enriches each image with complementary textual context at both local and global scopes. Local context goes beyond the surrounding text by incorporating sentences that are semantically related to the image, while global context provides a summary of the entire passage. We further introduce a multi-granularity design for the local context, allowing it to capture semantically relevant information at different levels of detail. Extensive experiments on the selected vision-centric dataset validate that CEMMKG is effective in leveraging contextual information to improve MMKG-based RAG performance. Moreover, its effectiveness across different MMKG-based RAG methods demonstrates its broad applicability.

cs.AI↗

arXiv2026

Expert-level vision-language foundation model for real-world radiology and comprehensive evaluation

Radiology is a vital and complex component of modern clinical workflow and covers many tasks. Recently, vision-language (VL) foundation models in medicine have shown potential in processing multimodal information, offering a unified solution for various radiology tasks. However, existing studies either pre-trained VL models on natural data or did not fully integrate vision-language architecture and pretraining, often neglecting the unique multimodal complexity in radiology images and their textual contexts. Additionally, their practical applicability in real-world scenarios remains underexplored. Here, we present RadFound, a large and open-source vision-language foundation model tailored for radiology, that is trained on the most extensive dataset of over 8.1 million images and 250,000 image-text pairs, covering 19 major organ systems and 10 imaging modalities. To establish expert-level multimodal perception and generation capabilities, RadFound introduces an enhanced vision encoder to capture intra-image local features and inter-image contextual information, and a unified cross-modal learning design tailored to radiology. To fully assess the models' capability, we construct a benchmark, RadVLBench, including radiology interpretation tasks like medical vision-language question-answering, as well as text generation tasks ranging from captioning to report generation. We also propose a human evaluation framework. When evaluated on the real-world benchmark involving three representative modalities, 2D images (chest X-rays), multi-view images (mammograms), and 3D images (thyroid CT scans), RadFound significantly outperforms other VL foundation models on both quantitative metrics and human evaluation. In summary, the development of RadFound represents an advancement in radiology generalists, demonstrating broad applicability potential for integration into clinical workflows.

cs.CV↗

arXiv2026

SlideGen: Collaborative Multimodal Agents for Scientific Slide Generation

Creating presentation slides from scientific papers is not simply a matter of summarizing paragraphs. A presenter is required to decide what story to tell, which figures and equations to highlight, and how to arrange them into pages that are visually clear rather than crowded or repetitive. The need to jointly reason over long contexts and layout-sensitive design makes paper-to-slide generation a uniquely challenging multimodal task. Most existing approaches, however, focus mainly on textual content selection, producing slides that often lack visual balance, narrative flow, or coherent integration of multimodal evidence. In this work, we introduce SlideGen, a collaborative vision-language multi-agent framework that coordinates narrative planning, multimodal grounding, and layout composition. SlideGen assigns specialized agents to outline the presentation structure, align supporting figures and tables with key claims, generate speaker notes, and compose editable PPTX slides through a diverse layout library. By refining layouts at the deck level, the system produces slide decks that are both faithful to the source paper and effective as presentations. To evaluate slide generation beyond text fidelity, we propose geometry-aware density (GAD), a metric that captures visual clutter, sparsity, and fragmentation, and shows strong agreement with human judgments. Evaluated across four complementary dimensions on our 200-paper benchmark, SlideGen consistently and significantly improves layout balance, content coverage, and text coherence, outperforming competitive baselines in paper-to-slide generation. Our findings suggest that effective slide generation requires multimodal design reasoning, and that agent collaboration offers a principled bridge between document understanding and scientific communication.

cs.AI↗

arXiv2026

The number of limit cycles of piecewise linear Liénard systems

For the planar Liénard differential system $\dot{x}=F(x)-y$, $\dot{y}=x$, where $F(x)$ is a piecewise linear function, Tonnelier (SIAM J. Appl. Math., 2002) conjectured that the maximum number of limit cycles of the system is $n$ when $F(x)$ has $n$ fold points and no jump points, and $2n$ when $F(x)$ has $n$ jump points and no fold points. This conjecture was confirmed by Llibre et al. (J. Nonlinear Sci., 2015) (resp. Chen et al. (J. London Math. Soc., 2026a)) when $F(x)$ has one fold point and no jump points (resp. two fold points). More recently, Chen et al. (J. London Math. Soc., 2026b) proved that the conjecture is correct when $F(x)$ has no fold points and one jump point. All other cases remain open. Here we verify that the lower bound for the maximum number of limit cycles of the system can be $n$ when $F(x)$ has only $n$ fold points, and $2n$ when $F(x)$ has only $n$ jump points, thereby confirming the lower bound part of Tonnelier's conjecture. Moreover, when $F(x)$ has $m$ jump points and $n-m$ fold points, $0\le m\le n$, we also show that the system can have $n+m=(n-m)+2m$ limit cycles. In addition, a complete classification of the {dynamics} near infinity for this class of systems is provided.

math.DS↗

arXiv2026

NaLA: A 3D Native LLM Layout Agent for High-quality 3D Scene Generation

Recently, Large Language Models (LLMs) have emerged as promising layout agents for 3D scene generation. Existing layout agents still suffer from implausible layout generation because most of them convert 3D assets and 3D layouts into textual descriptions as inputs and outputs, which involves severe information loss due to the modality gap between texts and 3D assets and 3D layouts. We propose NaLA, a native 3D LLM layout Agent for high-quality 3D scene generation by placing 3D assets in the scene. For the inputs, NaLA encodes 3D scene boundaries and 3D assets directly into the LLM, preserving fine-grained geometry and enabling explicit reasoning over relationships like collisions, surface supporting, and containment. To accurately output the positions and orientations of assets, NaLA adopts a coarse-to-fine prediction mechanism that first predicts discrete poses in an autoregressive manner and then refines the discrete poses with a continuous regression. Trained on diverse layout datasets, NaLA attains strong geometric perception and layout coherence. Experiments demonstrate that NaLA outperforms prior layout agents in both generation quality and inference efficiency, with comprehensive ablation studies to verify each component's effectiveness.

cs.CV↗

arXiv2026

On the Effect of Sampling Diversity in Scaling LLM Inference

Large language model (LLM) scaling inference is key to unlocking greater performance, and leveraging diversity has proven an effective way to enhance it. Motivated by the observed relationship between solution accuracy and meaningful response diversity, we systematically study the effect of prompt diversity in scaling inference. We theoretically explain why diversified sampling improves Best-of-$N$ scaling, showing that responses generated from diverse prompts after Best-of-$N$ selection exhibit significantly lower error rates than those produced from stationary prompts. Building on this analysis, we derive a diversity-fidelity trade-off principle, that guides the design of sampling strategies introducing diversity. From this guidance, we instantiate a family of effective perturbation styles. We theoretically and empirically characterize \textbf{when} diversified exploration remains effective, demonstrating that it works under a variety of conditions, and we further show that under majority voting, diversity may vanish. Finally, we systematically evaluate the effectiveness of sampling diversity and show that, when applied appropriately in different contexts, meaningful perturbations yield stronger, task-dependent gains as diversity increases. Overall, this work provides a systematic analysis that offers a theoretical and empirical foundation for understanding how sampling diversity affects LLM inference-time scaling.

cs.LG↗

arXiv2026

AirKey: Multimodal Acoustic-Assisted WiFi Sensing for Zero-Training Robust PIN Inference

Contactless keystroke inference via WiFi sensing highlights severe privacy threats, yet its real-world feasibility is hindered by two fundamental physical and deployment bottlenecks: the strict requirement for network privileges to acquire stable sensing streams, and the inherent "waveform fusion" ambiguity of pure WiFi signals during rapid, muscle-memory typing. To overcome these limitations, we propose AirKey, a novel cross-modal sensing framework that achieves highly stealthy, zero-training PIN eavesdropping. First, to bypass network deployment barriers, AirKey exploits fundamental IEEE 802.11 mechanisms to predictably elicit Acknowledgment (ACK) responses from unmodified target devices. By passively harvesting Channel State Information (CSI) from these ACKs using a low-cost microcontroller, AirKey secures a continuous spatial sensing stream entirely without network association. Crucially, to resolve the WiFi waveform fusion bottleneck, AirKey introduces a cross-modal complementarity mechanism. By utilizing lightweight acoustic signals as precise temporal anchors, the system robustly guides the segmentation of overlapping CSI trajectories. This joint spatiotemporal fusion strictly intersects CSI-derived spatial similarities with acoustic-guided inter-keystroke timing. Extensive real-world evaluations demonstrate that AirKey achieves over 4x higher accuracy than state-of-the-art unimodal zero-training schemes, successfully recovering device-unlock PINs within 6 attempts. Ultimately, this work exposes a critical vulnerability in contemporary smart interfaces, underscoring the severe privacy implications of ubiquitous multimodal sensing.

cs.CR↗

arXiv2026

RaG-Tree: Combining R-Tree and HNSW for Multi-Attribute Range Filtered Approximate Nearest Neighbor Search

Multi-attribute range-filtered approximate nearest neighbor search (MR-ANNS), which retrieves high-dimensional vectors satisfying multiple attribute constraints, is a fundamental operation in modern AI applications. Existing MR-ANNS indexes either exploit a single attribute for range localization or recursively partition objects along individual attributes, which may limit their ability to exploit attribute correlations for effective range pruning and attribute-vector correlations for efficient nearest-neighbor search. In this paper, we propose RaG-Tree, a unified index that couples an R-tree with partition-aware HNSW graphs for MR-ANNS. RaG-Tree leverages hierarchical R-tree partitions for effective range pruning and adapts the sparsity of each HNSW graph to the local vector distributions within its partition, enabling lightweight indexing and efficient query processing. To support efficient query processing and dynamic updates, we develop a cost-based adaptive search algorithm that minimizes unnecessary graph exploration, together with an efficient index maintenance mechanism for incrementally updating affected partition-aware HNSW graphs. Extensive experiments on three real-world datasets show that RaG-Tree achieves superior query performance over state-of-the-art baselines, while also providing lightweight indexing and fast incremental updates.

cs.DB↗

arXiv2026

QuantHarness: Price-Driven Multi-Agent LLMs for High-Frequency Trading

Recent advances in Large Language Models (LLMs) have shown remarkable capabilities in financial reasoning and market understanding. Multi-agent LLM frameworks such as TradingAgent and FINMEM augment these models to long-horizon investment tasks by leveraging fundamental and sentiment-based inputs for strategic decision-making. However, these approaches are ill-suited for the high-speed, precision-critical demands of High-Frequency Trading (HFT). HFT typically requires rapid, risk-aware decisions driven by structured, short-horizon signals, such as technical indicators, chart patterns, and trend features. These signals stand in sharp contrast to the long-horizon, text-driven reasoning that characterizes most existing LLM-based systems in finance. To bridge this gap, we introduce QuantHarness, the first multi-agent LLM framework explicitly designed for high-frequency algorithmic trading. The system decomposes trading into four specialized agents--Indicator, Pattern, Trend, and Risk--each equipped with domain-specific tools and structured reasoning capabilities to capture distinct aspects of market dynamics over short temporal windows. Extensive experiments across nine financial instruments, including Bitcoin and Nasdaq futures, demonstrate that QuantHarness consistently outperforms baseline methods, achieving higher predictive accuracy at both 1-hour and 4-hour trading intervals across multiple evaluation metrics. Our findings suggest that coupling structured trading signals with LLM-based reasoning provides a viable path for traceable, real-time decision systems in high-frequency financial markets.

cs.CE↗

arXiv2026

Boundary-Phase Control of Sequentially Addressed Trapped-Ion ZZ Interactions

Motion-mediated trapped-ion interactions commonly coordinate state-dependent forces on both target ions. Sequential optical access reduces the number of concurrent target channels but makes the relative phase between disjoint force windows a control variable. We derive a complex near-resonant description in which each window generates a displacement vector and ordered symplectic products between vectors on different ions produce the ZZ phase. Only relative boundary phases affect this area; a common phase shift is a gauge transformation. Building on the experimental precedent for alternating single-ion addressing, we develop matched-envelope phase and contrast controls that isolate this boundary-phase dependence without target-window overlap or hidden force in the dark gaps. The analysis separates phase generation from differential closure, projector-common motion, deterministic local-Z phases, spectator coupling, and control-parameter transfer. A conditional-Ramsey sequence gives continuous and reset contrasts of 0.998 and 0.996, with a reset-induced phase separation of 0.581 rad modulo $π/2$. In the representative comparison, sequential control uses fewer concurrent target channels but greater normalized force action than independently calibrated simultaneous control. All results are model-level estimates within the stated Lamb-Dicke, rotating-wave, and apparatus-input limits.

quant-ph↗

arXiv2026

Geometry-Resolved Projection of RF Imbalance to Ion Micromotion in a Same-Phase Dual-RF Blade Trap

Common-mode metrics of a high-$Q$ helical resonator do not determine the residual ion-side field in a dual-electrode drive. We combine a two-node differential RF model with single-electrode finite-element bases to obtain a computation-only, geometry-resolved projection for a same-phase blade trap. For a 3.5 pF external load per branch, the model gives a total effective branch capacitance of 7.640 pF and an HWHM-equivalent full branch-difference scale of 12.7 fF at $Q_{\mathrm{loaded}}=600$. The seven-segment geometry gives center and axial-RMS differential field coefficients of 640 V m$^{-1}$ and 635 V m$^{-1}$ per differential peak volt. A representative 10 fF mismatch with an effective 0.1 pF balance scale projects to 44.5/44.1 nm center/RMS $^{171}\mathrm{Yb}^{+}$ micromotion at 100 V common peak voltage. Supplementary thermal, bypass-admittance, and tested numerical cases characterize model sensitivity. All reported displacements are projections; no RF-bench or ion-side validation is claimed.

quant-ph↗

arXiv2026

Fate of dynamical quantum phase transitions from sudden quench to slow quench limit

Previous investigations of dynamical quantum phase transition (DQPT) have predominantly focused on sudden quench protocols. In this work, we systematically explore the fate of DQPT from the sudden to the slow quench regime. We establish that DQPT remains robust under slow quenching when the protocol crosses an equilibrium quantum phase transition. Conversely, we show that accidental DQPTs,which occur when the pre and post quench Hamiltonians reside in the same equilibrium phase, can be filtered out in the slow quench limit. This demonstrates that slow quenches can reveal the intrinsic connection between DQPT and equilibrium quantum phase transitions, thereby enabling DQPT to serve as a dynamical probe for identifying equilibrium quantum critical points. We illustrate these findings with examples from the XY chain, the Aubry-André model, and the trimer Su-Schrieffer-Heeger model.

quant-ph↗

arXiv2026

What Causes Performance Degradation in Cross-Subject EEG Classification?

Cross-subject Electroencephalography (EEG) classification typically achieves significantly lower performance than subject-dependent settings. Although this phenomenon has been widely observed in the literature, the underlying causes have not been systematically studied. In this paper, we design a series of controlled experiments to investigate the mechanisms behind the performance drop in cross-subject EEG classification across different EEG tasks. We show that the performance degradation can generally be attributed to two factors: inter-subject variability and shortcut learning. Specifically, multi-class-per-subject EEG classification tasks, such as motor imagery, emotion recognition, and ERP stimulus classification, are mainly affected by inter-subject variability, whereas single-class-per-subject EEG classification tasks, such as brain disease detection, are primarily influenced by shortcut learning based on subject-specific features. These findings provide new insights into the challenges of cross-subject EEG classification and emphasize the importance of appropriate evaluation protocols in EEG research. The code is available at https://github.com/DL4mHealth/EEG-Cross-Subject.

cs.CE↗

arXiv2026

Technical Report on the CVPR 2026@AdvML Workshop Challenge

Vision-language agents (VLAs) are increasingly used to interpret complex driving scenes and support safety-critical reasoning. This report presents the CVPR 2026@AdvML Workshop Challenge on adversarial multimodal attacks against autonomous-driving VLAs. Built on DriveLM-style multi-view visual question answering, the challenge represents each scene with six synchronized camera images and a structured collection of driving-related question-answer pairs. Participants generate adversarial images and suffix-only textual perturbations that induce model responses to deviate from reference answers while preserving image fidelity and limiting textual cost. The competition comprises two phases, with Phase II adding a hidden black-box model to assess transferability. We describe the task design, submission rules, evaluation protocol, and leaderboard results, and then examine five leading submissions for which technical reports were available. Across these reports, several recurring patterns emerge: image-side attacks are favored by the suffix penalty; scene-level, multi-view optimization is more effective than treating views in isolation; QA types and graph structure provide useful priors for allocating attack budget; feature-space objectives can improve black-box transfer; and typographic content embedded in camera images exposes a persistent vulnerability in driving VLAs. These findings provide a practical reference for future robustness evaluation and defense design in multimodal autonomous-driving systems.

cs.CV↗

arXiv2026

Inferring Inventory Dynamics from Supply Chain Networks: A Graph Learning Approach with Autonomous Validation

Supply-demand mismatch represents a fundamental challenge in supply chain management, yet its direct measurement remains particularly elusive for small and medium-sized enterprises (SMEs).These firms typically lack systematic inventory records, leaving labeled training data critically scarce. Conventional supervised learning methods rely heavily on labeled samples, rendering them ill-equipped to reliably validate firm-level predictions under such data-scarce conditions. To resolve this unlabeled-data dilemma, we develop a multi-agent semi-supervised inference framework that reframes the label-scarcity problem as a structured, collaborative task distributed across specialized agents. We first construct a production-function-constrained graph machine learning model that infers firm-level inventory changes directly from supply chain network topology. A dedicated econometric validation agent then concurrently loads five econometric models (spanning spatial spillovers, dynamic persistence, causal direction, shock transmission, and supply-demand forecasting) to generates structured economic evidence from complementary dimensions. An expert review agent synthesizes the structured econometric evidence and produces a unified consistency assessment by resolving cross-agent inconsistencies. Empirical results demonstrate stable predictive performance on inventory-change forecasting tasks. Multi-agent econometric validation further confirms that predicted inventory dynamics align closely with established economic theory in terms of causal structure and network transmission mechanisms. Critically, the proposed agent framework enables effective prediction verification even when ground-truth observations are unavailable.

stat.AP↗

arXiv2026

Tentative detection of circularly polarized bursty radio emissions from the HD 189733 exoplanetary system using NenuFAR beamformed observations

Observing auroral radio emission is one of the most promising methods for detecting exoplanetary magnetic fields, which provide valuable insights into planetary interiors, atmospheric properties, and potential habitability. The first hints of exoplanet auroral emission are starting to emerge. Recently, Zhang et al. (2025) reported a detection at 50 MHz of a circularly polarized bursty emission from the HD 189733 exoplanetary system using NenuFAR low-frequency imaging observations. The source of the emission is still unknown and may be caused by planetary auroral emissions, star-planet interactions, stellar activity, or the M-dwarf stellar companion. In this study, we analyze beamformed observations from NenuFAR of HD 189733 taken simultaneously during the previously detected burst. This dataset allows for an independent verification of the detected burst with a different backend and processing steps. Using the BOREALIS data reduction pipeline, we tentatively detect circularly polarized bursty emission ($\sim$10$σ$) from HD 189733 $\sim$1 hour before the burst found from the imaging observations. However, some uncertainty remains on whether our detected signal is astrophysical in nature due to excess correlated noise. Assuming an astrophysical origin, our observed characteristics are most consistent with a planetary origin, but stellar emission cannot be completely ruled. Therefore, more low-frequency radio observations are needed to confirm the astrophysical nature of our signal and to search for periodicity in the radio signal from HD 189733 to determine the true cause of the emission. These observations are ongoing. Our study highlights the power of simultaneous beamformed and imaging observations in the search for radio emission from exoplanets.

astro-ph.EP↗

arXiv2026

Quantum Dot Moiré from Crossed MoS2 Nanoribbons

Twisted atomically thin layers have attracted much attention for Moiré potential and correlated quantum phenomena. However, existing Moiré superlattices have largely been limited to extensive wavefunction without lateral confinement. Here we introduce a new platform where 1D nanoribbons of 2D MoS2 grown by vapor deposition can be easily superposed at various angles from stacking and transferring, to form Moiré quantum dots at their intersections with unique exciton physics. Angle-dependent Moiré intersections show enhanced exciton emission at commensurate angle 22 deg, which demonstrates faster relaxation at the cryogenic temperature. A size-dependent study further exhibits a reduced exciton energy and soften out-of-plane interlayer coupling for smaller Moiré areas. Our results reveal exciton physics turnability via precise overlapping of 1D nanoribbons.

cond-mat.mtrl-sci↗