arXiv subjects

Gang Xu

Publications and source records attributed to Gang Xu.

At least 37 records · Page 2Linked to original sources

arXiv2026

ORACLE: Anticipating Scams from Partial Trajectories in Streaming App Usage

Smartphone scams are increasingly prevalent and typically manifest as multi-stage, cross-application processes with gradually emerging intent. Effective intervention thus requires anticipating scams before the intent becomes explicit. This is inherently challenging, as decisions must rely on partial trajectories with temporally distributed evidence. In this paper, we propose \textbf{ORACLE} Online Reasoning for Anticipating Cross-temporal Latent thrEats, the first agentic framework for early scam anticipation from \textit{streaming app-usage} trajectories. To support this setting, we curate a real-world long-horizon benchmark of streaming app-usage trajectories, covering 12 scam types, spanning extended periods (15 days on average), involving diverse applications (95 apps), and interleaving normal and scam behaviors. To address fragmented evidence, we introduce a self-evolving context manager that adaptively consolidates entity-centric interactions over time, enabling more effective reconstruction of cross-temporal evidence from partial observations. To enhance sensitivity to latent early-stage signals, we propose an on-policy self-distillation scheme in which a teacher model, conditioned on summarized anti-scam reflections and clues by skills, supervises a student model without access to such reflections. This scheme thereby distills evidence-informed knowledge and improves recognition of emerging fraud patterns from partial trajectories. Experiments show that \method{} consistently improves early scam anticipation, yielding timely warnings while reducing false alerts in realistic streaming scenarios.

cs.LG↗

arXiv2026

Risk Awareness Injection: Calibrating Vision-Language Models for Safety without Compromising Utility

Vision language models (VLMs) extend the reasoning capabilities of large language models (LLMs) to cross-modal settings, yet remain highly vulnerable to multimodal jailbreak attacks. Existing defenses predominantly rely on safety fine-tuning or aggressive token manipulations, incurring substantial training costs or significantly degrading utility. Recent research shows that LLMs inherently recognize unsafe content in text, and the incorporation of visual inputs in VLMs frequently dilutes risk-related signals. Motivated by this, we propose Risk Awareness Injection (RAI), a lightweight and training-free framework for safety calibration that restores LLM-like risk recognition by amplifying unsafe signals in VLMs. Specifically, RAI constructs an Unsafe Prototype Subspace from language embeddings and performs targeted modulation on selected high-risk visual tokens, explicitly activating safety-critical signals within the cross-modal feature space. This modulation restores the model's LLM-like ability to detect unsafe content from visual inputs, while preserving the semantic integrity of original tokens for cross-modal reasoning. Extensive experiments across multiple jailbreak and utility benchmarks demonstrate that RAI substantially reduces attack success rate without compromising task performance.

cs.AI↗

arXiv2026

RS-WorldModel: a Unified Model for Remote Sensing Understanding and Future Sense Forecasting

Remote sensing world models aim to both explain observed changes and forecast plausible futures, two tasks that share spatiotemporal priors. Existing methods, however, typically address them separately, limiting cross-task transfer. We present RS-WorldModel, a unified world model for remote sensing that jointly handles spatiotemporal change understanding and text-guided future scene forecasting, and we build RSWBench-1.1M, a 1.1 million sample dataset with rich language annotations covering both tasks. RS-WorldModel is trained in three stages: (1) Geo-Aware Generative Pre-training (GAGP) conditions forecasting on geographic and acquisition metadata; (2) synergistic instruction tuning (SIT) jointly trains understanding and forecasting; (3) verifiable reinforcement optimization (VRO) refines outputs with verifiable, task-specific rewards. With only 2B parameters, RS-WorldModel surpasses open-source models up to 120$ \times $ larger on most spatiotemporal change question-answering metrics. It achieves an FID of 43.13 on text-guided future scene forecasting, outperforming all open-source baselines as well as the closed-source Gemini-2.5-Flash Image (Nano Banana).

cs.AI↗

arXiv2026

Yo'City: Personalized and Boundless 3D Realistic City Scene Generation via Self-Critic Expansion

Realistic 3D city generation is fundamental to a wide range of applications, including virtual reality and digital twins. However, most existing methods rely on training a single diffusion model, which limits their ability to generate personalized and boundless city-scale scenes. In this paper, we present Yo'City, a novel agentic framework that enables user-customized and infinitely expandable 3D city generation by leveraging the reasoning and compositional capabilities of off-the-shelf large models. Specifically, Yo'City first conceptualizes the city through a top-down planning strategy that defines a hierarchical "City-District-Grid" structure. The Global Planner determines the overall layout and potential functional districts, while the Local Designer further refines each district with detailed grid-level descriptions. Subsequently, the grid-level 3D generation is achieved through a "produce-refine-evaluate" isometric image synthesis loop, followed by image-to-3D generation. To simulate continuous city evolution, Yo'City further introduces a user-interactive, relationship-guided expansion mechanism, which performs scene graph-based distance- and semantics-aware layout optimization, ensuring spatially coherent city growth. To comprehensively evaluate our method, we construct a diverse benchmark dataset and design six multi-dimensional metrics that assess generation quality from the perspectives of semantics, geometry, texture, and layout. Extensive experiments demonstrate that Yo'City consistently outperforms existing state-of-the-art methods across all evaluation aspects.

cs.CV↗

arXiv2026

BioChemInsight: An Online Platform for Automated Extraction of Chemical Structures and Activity Data from Patents

The automated extraction of chemical structures and their corresponding bioactivity data is essential for accelerating drug discovery and enabling data-driven research. Current optical chemical structure recognition tools lack the capability to autonomously link molecular structures with their bioactivity profiles, posing a significant bottleneck in structure-activity relationship analysis. To address this, we present BioChemInsight, an open-source pipeline that integrates DECIMER Segmentation with MolNexTR for chemical structure recognition, GLM-4.5V for compound identifier association, and PaddleOCR combined with GLM-4.6 for bioactivity extraction and unit normalization. We evaluated BioChemInsight on 181 patents covering 15 therapeutic targets. The system achieved an average extraction accuracy of above 90% across three key tasks: chemical structure recognition, bioactivity data extraction, and compound identifier association. Our analysis indicates that the chemical space covered by patents is largely complementary to that contained in established public database ChEMBL. Consequently, by enabling systematic patent mining, BioChemInsight provides access to chemical information underrepresented in ChEMBL. This capability expands the landscape of explorable compound-target interactions, enriches the data foundation for quantitative structure-activity relationship modeling and targeted screening, and reduces data preprocessing time from weeks to hours. BioChemInsight is available at https://github.com/dahuilangda/BioChemInsight.

q-bio.QM↗

arXiv2026

Symmetry-Broken Cavity Solitons and Collective Polarization Conformity in Fabry-Perot Kerr Resonators

We report on the experimental generation of polarization symmetry-broken cavity solitons (CSs) in a passive, fiber-based, coherently-driven, Fabry-Perot (FP) Kerr resonator. Polarization resolved measurements reveal the spontaneous transition of initially symmetric CSs into asymmetrical vectorial states, triggered by a cross-phase modulation-induced polarization bifurcation. Most notably, due to counter-propagation of light occurring in FP resonators, we unveil a collective polarization conformity effect, whereby multiple CSs circulating in the cavity converge to the same asymmetric polarization state once their number exceeds a certain threshold. These results demonstrate that Fabry-Perot resonators support novel collective soliton dynamics that are absent in ring architectures.

physics.optics↗

arXiv2026

Uniaxial stress enhanced anisotropic magnetoresistance and superconductivity in the kagome superconductor LaRu$_{3}$Si$_{2}$

Elucidating the role of the kagome electronic structure in determining the various quantum ground states is of fundamental importance. In this work, we employ in-plane uniaxial stress as a tuning parameter to probe the electronic structure and its impact on the superconducting and normal-state properties of the kagome superconductor LaRu$_{3}$Si$_{2}$, combining magnetotransport measurements with first-principles calculations. We identify a pronounced anisotropy in both the upper critical field and the normal-state magnetoresistance, indicating strong electronic anisotropy despite the three-dimensional crystal structure. Furthermore, we find that the superconducting transition temperature $T_{\rm c}$ increases under in-plane stress applied within the kagome plane, although the enhancement is modest, reaching approximately 0.3 K at 0.6 GPa. Furthermore, the absolute magnetoresistance exhibits a pronounced increase from about 22${\%}$ at zero stress to 35${\%}$ at 0.6 GPa, indicating a substantial modification of the normal state above $T_{\rm c}$. Previous studies have reported time-reversal-symmetry (TRS) breaking below a temperature scale that coincides with the onset of magnetoresistance. The simultaneous enhancement of both $T_{\rm c}$ and magnetoresistance under stress therefore suggests a positive correlation between superconductivity and normal-state electronic and magnetic properties in LaRu$_{3}$Si$_{2}$. Detailed calculations demonstrate that stress-induced changes in $T_{\rm c}$ arise from the joint evolution of the total density of states and the flat band, whereas the large magnetoresistance enhancement is dominated by the stress-driven downward shift of the Ru $dz^{2}$ kagome flat band.

cond-mat.supr-con↗

arXiv2026

LongCat-Flash-Thinking-2601 Technical Report

We introduce LongCat-Flash-Thinking-2601, a 560-billion-parameter open-source Mixture-of-Experts (MoE) reasoning model with superior agentic reasoning capability. LongCat-Flash-Thinking-2601 achieves state-of-the-art performance among open-source models on a wide range of agentic benchmarks, including agentic search, agentic tool use, and tool-integrated reasoning. Beyond benchmark performance, the model demonstrates strong generalization to complex tool interactions and robust behavior under noisy real-world environments. Its advanced capability stems from a unified training framework that combines domain-parallel expert training with subsequent fusion, together with an end-to-end co-design of data construction, environments, algorithms, and infrastructure spanning from pre-training to post-training. In particular, the model's strong generalization capability in complex tool-use are driven by our in-depth exploration of environment scaling and principled task construction. To optimize long-tailed, skewed generation and multi-turn agentic interactions, and to enable stable training across over 10,000 environments spanning more than 20 domains, we systematically extend our asynchronous reinforcement learning framework, DORA, for stable and efficient large-scale multi-environment training. Furthermore, recognizing that real-world tasks are inherently noisy, we conduct a systematic analysis and decomposition of real-world noise patterns, and design targeted training procedures to explicitly incorporate such imperfections into the training process, resulting in improved robustness for real-world applications. To further enhance performance on complex reasoning tasks, we introduce a Heavy Thinking mode that enables effective test-time scaling by jointly expanding reasoning depth and width through intensive parallel thinking.

cs.AI↗

arXiv2026

Unexpected type-II multiferroic phase in GdMnO3 under high magnetic fields

Perovskite manganites with small A-site ions, as the first and canonical branch of type-II multiferroics, are ideal systems to exhibit magnetism-induced ferroelectricity. Despite their established magnetoelectric phase diagrams under low magnetic fields, here an unidentified phase with a large magnetism-induced polarization (up to 1500 μC/m2) is revealed in GdMnO3 under high magnetic fields up to 60 T. Based on multiprobe experiments, a complete phase diagram is constructed with successive polar-nonpolar-polar-nonpolar transitions. Such a nonmonotonic evolution is well mimicked by model simulation, while the spin-lattice coupling is the key ingredient for the reentrant ferroelectric phase.

cond-mat.mtrl-sci↗

arXiv2025

Multi-peak vector soliton families in defocusing Kerr resonators

We report the existence of multi-peaked vector soliton families in normally dispersive passive Kerr resonators. Through cross-phase modulation between two orthogonal polarization components, each peak becomes tightly interlocked, enabling robust localization of the entire wave packet in defocusing cavities. Analysis using snakes-and-ladder diagrams demonstrates the diversity of these vector soliton families, which include dark-bright multi-peak solitons, flat-topped solitons, and modulation instability patterns, among others. Furthermore, stability analysis based on the coupled Lugiato-Lefever equations reveals that specific combinations of parameters can sustain stable vector cavity solitons, whose peak numbers can be continuously tuned by adding appropriate perturbations. These findings significantly expand the scope of soliton dynamics and optical frequency comb generation in pumped-dissipative systems, independent of dispersion conditions.

physics.optics↗

arXiv2025

LongCat-Flash-Omni Technical Report

We introduce LongCat-Flash-Omni, a state-of-the-art open-source omni-modal model with 560 billion parameters, excelling at real-time audio-visual interaction. By adopting a curriculum-inspired progressive training strategy that transitions from simpler to increasingly complex modality sequence modeling tasks, LongCat-Flash-Omni attains comprehensive multimodal capabilities while maintaining strong unimodal capability. Building upon LongCat-Flash, which adopts a high-performance Shortcut-connected Mixture-of-Experts (MoE) architecture with zero-computation experts, LongCat-Flash-Omni integrates efficient multimodal perception and speech reconstruction modules. Despite its immense size of 560B parameters (with 27B activated), LongCat-Flash-Omni achieves low-latency real-time audio-visual interaction. For training infrastructure, we developed a modality-decoupled parallelism scheme specifically designed to manage the data and model heterogeneity inherent in large-scale multimodal training. This innovative approach demonstrates exceptional efficiency by sustaining over 90% of the throughput achieved by text-only training. Extensive evaluations show that LongCat-Flash-Omni achieves state-of-the-art performance on omni-modal benchmarks among open-source models. Furthermore, it delivers highly competitive results across a wide range of modality-specific tasks, including text, image, and video understanding, as well as audio understanding and generation. We provide a comprehensive overview of the model architecture design, training procedures, and data strategies, and open-source the model to foster future research and development in the community.

cs.MM↗

arXiv2025

Introducing LongCat-Flash-Thinking: A Technical Report

We present LongCat-Flash-Thinking, an efficient 560-billion-parameter open-source Mixture-of-Experts (MoE) reasoning model. Its advanced capabilities are cultivated through a meticulously crafted training process, beginning with long Chain-of-Thought (CoT) data cold-start and culminating in large-scale Reinforcement Learning (RL). We first employ a well-designed cold-start training strategy, which significantly enhances the reasoning potential and equips the model with specialized skills in both formal and agentic reasoning. Then, a core innovation is our domain-parallel training scheme, which decouples optimization across distinct domains (e.g., STEM, Code, Agentic) and subsequently fuses the resulting expert models into a single, nearly Pareto-optimal model. This entire process is powered by our Dynamic ORchestration for Asynchronous rollout (DORA) system, a large-scale RL framework that delivers a greater than threefold training speedup over synchronous methods on tens of thousands of accelerators. As a result, LongCat-Flash-Thinking achieves state-of-the-art performance among open-source models on a suite of complex reasoning tasks. The model exhibits exceptional efficiency in agentic reasoning, reducing average token consumption by 64.5% (from 19, 653 to 6, 965) on AIME-25, without degrading task accuracy. We release LongCat-Flash-Thinking to promote further advances in reasoning systems and agentic AI research.

cs.AI↗

arXiv2025

Multiple Imputation for Small, Extremely High Efficacy Clinical Trials with Binary Endpoints

There has been an increasing interest in using cell and gene therapy (CGT) to treat/cure difficult diseases. The hallmark of CGT trials are the small sample size and extremely high efficacy. Due to the innovation and novelty of such therapies, when there is missing data, more scrutiny is exercised, and regulators often request for missing data handling strategy when missing data occurs. Often, multiple imputation (MI) will be used. MI for continuous endpoint is well established but literature of MI for binary endpoint is lacking. In this work, we compare and develop 3 new methods to handle missing data using MI for binary endpoints when the sample size is small and efficacy extremely high. The parameter of interest is population proportion of success. We show that our proposed methods performed well and produced good 95% coverage. We also applied our methods to an actual clinical study, the Clinical Islet Transplantation (CIT) Protocol 07, conducted by National Institutes of Health (NIH).

stat.AP↗

arXiv2025

Rich unconventional Hall effects in a single quasi-kagome Kondo Weyl semimetal candidate Ce$_3$TiSb$_5$

It is generally believed that electronic correlation, geometric frustration, and topology, \textit{individually}, can facilitate the emergence of various intriguing properties that have attracted a broad audience for both fundamental research and potential applications. Here, we report a series of unconventional Hall effects observed in a \textit{single} compound - quasi-kagome Kondo Weyl semimetal candidate Ce$_3$TiSb$_5$. In the paramagnetic phase, signature of dynamic $c$-$f$ hybridization is revealed by a reduction of anomalous Hall effect and is connected to frustration-promoted incoherent Kondo scattering. A large topological Hall effect exceeding 0.2 $μΩ\cdot$cm is found at low temperatures, which should be ascribed to the non-collinear magnetic texture. In addition, a peculiar loop-shaped Hall effect with switching chirality is also seen, which is inferred to be associated with magnetic domain walls that pin history-dependent spin chirality and / or Fermi-arc surface states projected from the in-gap Weyl nodes. These exotic results place Ce$_3$TiSb$_5$ in a regime of highly-frustrated antiferromagnetic dense Kondo lattice with a nontrivial topology on an ``extended" global phase diagram, and highlight the interplay among electronic correlation, geometric frustration and topology.

cond-mat.str-el↗

arXiv2025

LongCat-Flash Technical Report

We introduce LongCat-Flash, a 560-billion-parameter Mixture-of-Experts (MoE) language model designed for both computational efficiency and advanced agentic capabilities. Stemming from the need for scalable efficiency, LongCat-Flash adopts two novel designs: (a) Zero-computation Experts, which enables dynamic computational budget allocation and activates 18.6B-31.3B (27B on average) per token depending on contextual demands, optimizing resource usage. (b) Shortcut-connected MoE, which enlarges the computation-communication overlap window, demonstrating notable gains in inference efficiency and throughput compared to models of a comparable scale. We develop a comprehensive scaling framework for large models that combines hyperparameter transfer, model-growth initialization, a multi-pronged stability suite, and deterministic computation to achieve stable and reproducible training. Notably, leveraging the synergy among scalable architectural design and infrastructure efforts, we complete model training on more than 20 trillion tokens within 30 days, while achieving over 100 tokens per second (TPS) for inference at a cost of \$0.70 per million output tokens. To cultivate LongCat-Flash towards agentic intelligence, we conduct a large-scale pre-training on optimized mixtures, followed by targeted mid- and post-training on reasoning, code, and instructions, with further augmentation from synthetic data and tool use tasks. Comprehensive evaluations demonstrate that, as a non-thinking foundation model, LongCat-Flash delivers highly competitive performance among other leading models, with exceptional strengths in agentic tasks. The model checkpoint of LongCat-Flash is open-sourced to foster community research. LongCat Chat: https://longcat.ai Hugging Face: https://huggingface.co/meituan-longcat GitHub: https://github.com/meituan-longcat

cs.CL↗

arXiv2025

5d orbital Induced Room Temperature Quantum Anomalous Hall Effect in TbCl

Following the experimental realization of Quantum anomalous Hall (QAH) effect in thin films of chromium-doped (Bi,Sb)$_2$Te$_3$, enhancing the work temperature of QAH effect has emerged as a significant and challenging task. Here we demonstrate monolayer TbCl as a promising candidate to realize the room temperature QAH effect. Using DFT+U method, double checked by HSE06 and DMFT calculations, we identify the Hall conductivity $G = -e^2/h$ per layer in three-dimensional ferromagnetic insulator TbCl, which is a weakly stacking of QAH layers. The monolayer TbCl inherits the magnetic and topological properties, exhibiting the QAH effect with Chern number $C$=-1. The large topological band gap reaches 42.8 meV, which is beyond room temperatue. The extended 5$d$ electrons lead to sizable exchange and superexchange interactions, resulting in a high Curie temperature $T_c$$\sim$457K. All these features demonstrate that monolayer TbCl will provide an ideal platform to realize the room temperature QAH effect.

cond-mat.mtrl-sci↗

arXiv2025

TextSplat: Text-Guided Semantic Fusion for Generalizable Gaussian Splatting

Recent advancements in Generalizable Gaussian Splatting have enabled robust 3D reconstruction from sparse input views by utilizing feed-forward Gaussian Splatting models, achieving superior cross-scene generalization. However, while many methods focus on geometric consistency, they often neglect the potential of text-driven guidance to enhance semantic understanding, which is crucial for accurately reconstructing fine-grained details in complex scenes. To address this limitation, we propose TextSplat--the first text-driven Generalizable Gaussian Splatting framework. By employing a text-guided fusion of diverse semantic cues, our framework learns robust cross-modal feature representations that improve the alignment of geometric and semantic information, producing high-fidelity 3D reconstructions. Specifically, our framework employs three parallel modules to obtain complementary representations: the Diffusion Prior Depth Estimator for accurate depth information, the Semantic Aware Segmentation Network for detailed semantic information, and the Multi-View Interaction Network for refined cross-view features. Then, in the Text-Guided Semantic Fusion Module, these representations are integrated via the text-guided and attention-based feature aggregation mechanism, resulting in enhanced 3D Gaussian parameters enriched with detailed semantic cues. Experimental results on various benchmark datasets demonstrate improved performance compared to existing methods across multiple evaluation metrics, validating the effectiveness of our framework. The code will be publicly available.

cs.CV↗

arXiv2025

PVChat: Personalized Video Chat with One-Shot Learning

Video large language models (ViLLMs) excel in general video understanding, e.g., recognizing activities like talking and eating, but struggle with identity-aware comprehension, such as "Wilson is receiving chemotherapy" or "Tom is discussing with Sarah", limiting their applicability in smart healthcare and smart home environments. To address this limitation, we propose a one-shot learning framework PVChat, the first personalized ViLLM that enables subject-aware question answering (QA) from a single video for each subject. Our approach optimizes a Mixture-of-Heads (MoH) enhanced ViLLM on a synthetically augmented video-QA dataset, leveraging a progressive image-to-video learning strategy. Specifically, we introduce an automated augmentation pipeline that synthesizes identity-preserving positive samples and retrieves hard negatives from existing video corpora, generating a diverse training dataset with four QA types: existence, appearance, action, and location inquiries. To enhance subject-specific learning, we propose a ReLU Routing MoH attention mechanism, alongside two novel objectives: (1) Smooth Proximity Regularization for progressive learning through exponential distance scaling and (2) Head Activation Enhancement for balanced attention routing. Finally, we adopt a two-stage training strategy, transitioning from image pre-training to video fine-tuning, enabling a gradual learning process from static attributes to dynamic representations. We evaluate PVChat on diverse datasets covering medical scenarios, TV series, anime, and real-world footage, demonstrating its superiority in personalized feature understanding after learning from a single video, compared to state-of-the-art ViLLMs.

cs.CV↗