arXiv subjects

Xin Wu

Publications and source records attributed to Xin Wu.

At least 19 recordsLinked to original sources

arXiv2026

SuppreSensing: Expert-Guided Feature Recalibration and Discrepancy Augmentation for Multimodal Object Detection

Multimodal object detection in remote sensing faces challenges due to semantic heterogeneity and modality-specific noise interference. To this end, we propose SuppreSensing, which reformulates multimodal fusion as a selective collaboration process that jointly models shared information and modality-specific cues. SuppreSensing first designs an Expert-driven Multimodal Feature Recalibration (EMFR) module, which reformulates shared-consensus extraction as an input-adaptive multi-expert selection process to alleviate the symmetry trap in multimodal fusion. Complementing this, a modality-specific attribute augmentation strategy is employed to enhance specific modality features by modeling bidirectional discrepancy patterns, mitigating cross-modal heterogeneity. Furthermore, we propose an Expert-driven Customized Feature Purification (ECFP) module based on a "specialized inspection-comprehensive analysis-diagnostic update" physical examination paradigm to iteratively filter redundancies and reinforce task-relevant semantics. Extensive experiments on the DroneVehicle and VEDAI datasets demonstrate that SuppreSensing achieves state-of-the-art detection performance. Cross-domain evaluations on natural scene datasets (FLIR and LLVIP) further validate its superior robustness and generalization capability across diverse environmental conditions.

cs.CV↗

arXiv2026

Complete Trip: A Linked Multimodal Human Mobility Dataset

Human mobility data have become fundamental to research across transportation, public health, urban science, and disaster resilience. However, existing mobility datasets typically capture only isolated aspects of travel behavior and rarely provide linked multimodal journeys together with network-level route representations and population-level inference. Here we present Complete Trip, a mobility dataset that reconstructs linked multimodal travel behavior from passively collected smartphone location-based services (LBS) data. The first released implementation covers six counties in Utah throughout 2020 and represents journeys across car, bus, rail, and active transportation through a four-stage workflow consisting of trip identification, mode imputation, route reconstruction, and trip linking. Complete Trip preserves journey-level relationships by linking sequential travel segments where multiple segments belong to the same travel episode, provides network-based route representations on digital transportation networks, and supports population-level analyses through statistically calibrated expansion weights. By providing a representation of linked multimodal human mobility, Complete Trip enables reproducible research across transportation, public health, urban science, disaster resilience, and related fields.

cs.CY↗

arXiv2026

MedRealMM: A Real-World Multimodal Benchmark for Chinese Online Medical Consultation

Large language models (LLMs) are increasingly deployed in online medical consultation, yet existing benchmarks remain poorly aligned with real clinical practice. Many rely on synthetic conversations or patient simulators, omit patient-uploaded medical images, or evaluate open-ended clinical responses using multiple-choice or lexical-overlap metrics that poorly reflect clinical quality. We introduce \textbf{MedRealMM}, a large-scale benchmark for multimodal online medical consultation built from de-identified patient-doctor interactions collected from a nationwide Chinese internet hospital. MedRealMM uses a Multimodal Clinical Challenge Point (MCCP) extraction framework to identify clinically demanding moments in authentic consultation trajectories and converts each into a standardized next-response generation task while preserving the preceding text-image context. Each instance is paired with a case-specific rubric refined by physicians that rewards clinically desirable behaviors and penalizes unsafe, unsupported, or contradictory responses. The current release contains 5,620 real-world multimodal cases spanning 64 clinical departments. We evaluate 19 general-purpose and medical-specialized LLMs, including text-only and multimodal systems. Our results show that image information is critical for reliable clinical performance and that current frontier models remain below the online physician response. Although some frontier models satisfy as many or more positive clinical criteria than physicians, they trigger more negative criteria, indicating that safety-sensitive error avoidance remains a central bottleneck. MedRealMM offers a realistic and reproducible benchmark for evaluating multimodal medical reasoning in real-world online consultation. The dataset will be publicly available on Hugging Face at https://huggingface.co/datasets/jdh-algo/MedRealMM.

cs.AI↗

arXiv2026

Truth or Sophistry? LoFa: A Benchmark for LLM Robustness Against Logical Fallacies

Large Language Models (LLMs) exhibit strong semantic capabilities, yet their resilience to manipulative linguistic patterns such as logical fallacies remains underexplored. Prior work has primarily examined whether LLMs can identify or classify fallacies, leaving their robustness against fallacious persuasion insufficiently studied. To address this gap, we introduce LoFa (Logical Fallacy), a comprehensive benchmark for evaluating LLM robustness against fallacies. LoFa is constructed through a multi-agent pipeline that pairs factual questions with fallacious arguments, and is accompanied by a multi-round debate framework for assessing model resilience under sustained adversarial persuasion. To disentangle fallacy robustness from a model's inherent knowledge limitations, we further propose Logical Fallacy Resistance at k (LFR@k), a metric that quantifies resistance to fallacious attacks. Experiments show that LLMs exhibit varying levels of robustness across different fallacy types, revealing distinct vulnerability profiles among models.

cs.CL↗

arXiv2026

CDPR: Counterfactual Advantage-based Credit Assignment for Cost-Aware Sequential Medical Diagnosis

Clinical diagnosis is a step-by-step, cost-aware process: a physician orders examinations one at a time, observes the results, and updates the diagnosis before reaching a final conclusion. Most medical language models instead treat diagnosis as a one-pass classification task and ignore the trade-off between a test's value and its cost. We model diagnosis as a cost-aware sequential decision process and train the policy with reinforcement learning. The main difficulty is credit assignment: the only reliable signal comes once at the end of a long trajectory, so it scores a wasteful workup the same as an efficient one. We propose CDPR (Counterfactual Diagnostic Process Reward), which needs no expert labels and no learned critic. CDPR first finds the states where the policy hesitates, using the uncertainty of its action distribution, and then scores the chosen action by its advantage over the alternatives the policy itself would consider, estimated with short rollouts under a utility that balances correctness against test count, cost, and infeasible requests. A rollout cache reuses within-batch trajectories to keep the cost low. We integrate CDPR into GRPO and test it on one in-domain (MIMIC-IV) and two out-of-domain (ClinicalBench and a private hospital dataset) benchmarks. CDPR improves diagnostic accuracy while clearly reducing the number and cost of examinations.

cs.AI↗

arXiv2026

SMADE-IE: Sparse Multi-Agent Framework with Evidence-Driven Debate for Zero-Shot Information Extraction

Zero-shot information extraction (IE) with large language models (LLMs) has attracted increasing attention due to its flexibility in adapting to new schemas and domains without task-specific training. Existing approaches mainly rely on monolithic prompting, each-type prompting, or multi-agent debate. However, monolithic prompting often suffers from boundary and type errors, while each-type prompting and multi-agent debate introduce cross-type conflicts, redundant agent interactions, and substantial token overhead. To address these challenges, we propose SMADE-IE, a sparse and evidence-driven multi-agent framework for zero-shot IE. SMADE-IE first employs an Adaptive Mode Selector to dynamically route inputs into either a lightweight Global Extraction Mode or a Type-Centric Extraction Mode, reducing unnecessary type selection and reasoning noise. For conflicting predictions, we further introduce an Evidence-Driven Debate mechanism that structures arguments into Toulmin-style components and performs confidence aggregation through external evidence scoring and Bayesian updates. Experimental results on 9 benchmark datasets across NER, RE, and JERE tasks show that SMADE-IE consistently outperforms existing zero-shot IE baselines while also improving token efficiency through sparse agent selection and early-stopping debate.

cs.CL↗

arXiv2026

DockSmith: Scaling Reliable Coding Environments via an Agentic Docker Builder

Reliable Docker-based environment construction is a dominant bottleneck for scaling execution-grounded training and evaluation of software engineering agents. We introduce DockSmith, a specialized agentic Docker builder designed to address this challenge. DockSmith treats environment construction not only as a preprocessing step, but as a core agentic capability that exercises long-horizon tool use, dependency reasoning, and failure recovery, yielding supervision that transfers beyond Docker building itself. DockSmith is trained on large-scale, execution-grounded Docker-building trajectories produced by a SWE-Factory-style pipeline augmented with a loop-detection controller and a cross-task success memory. Training a 30B-A3B model on these trajectories achieves open-source state-of-the-art performance on Multi-Docker-Eval, with 39.72% Fail-to-Pass and 58.28% Commit Rate. Moreover, DockSmith improves out-of-distribution performance on SWE-bench Verified, SWE-bench Multilingual, and Terminal-Bench 2.0, demonstrating broader agentic benefits of environment construction.

cs.AI↗

arXiv2026

Design and preliminary performance study of the broad-band spectrometer detector for POLAR-2

POLAR-2, the successor of the POLAR experiment aboard China's Tiangong-2 space lab, is set to be deployed on the China Space Station. The POLAR-2 mission aims to conducting high-precision polarization measurements of high-energy transients with a primary focus on Gamma-Ray Bursts (GRBs), following POLAR's pioneering accurate polarization measurements of GRB prompt emission. One of the key advancements in POLAR-2 is the inclusion of a dedicated Broad-band Spectrometer Detector (BSD) instrument, designed to provide precise measurements of GRB location and spectral parameters, which are critical inputs for accurate polarization analysis of POLAR-2's dedicated High-energy Polarimetry Detector (HPD), which is made of plastic scintillator bars array. BSD employs a coded-aperture mask imaging technique and pixelated GAGG scintillation crystals, offering a wide half-coded field of view of ~132° x 125° and an operational energy range of 10-1000 keV. Simulation results indicate that the instrument can achieve a localization accuracy of approximately 1.5° for faint GRBs similar to GRB 170817A, satisfying the core requirements of GRB polarimetry with HPD. BSD also has moderate capability for GRB polarimetry, particularly at several hundred keV energy. This paper outlines the preliminary design of BSD and presents an overall evaluation of its expected scientific performance, based on extensive Monte Carlo simulations and preliminary ground-based calibration tests.

astro-ph.IM↗

arXiv2026

M100: An Orchestrated Dataflow Architecture Powering General AI Computing

As deep learning-based AI technologies gain momentum, the demand for general-purpose AI computing architectures continues to grow. While GPGPU-based architectures offer versatility for diverse AI workloads, they often fall short in efficiency and cost-effectiveness. Various Domain-Specific Architectures (DSAs) excel at particular AI tasks but struggle to extend across broader applications or adapt to the rapidly evolving AI landscape. M100 is Li Auto's response: a performant, cost-effective architecture for AI inference in Autonomous Driving (AD), Large Language Models (LLMs), and intelligent human interactions, domains crucial to today's most competitive automobile platforms. M100 employs a dataflow parallel architecture, where compiler-architecture co-design orchestrates not only computation but, more critically, data movement across time and space. Leveraging dataflow computing efficiency, our hardware-software co-design improves system performance while reducing hardware complexity and cost. M100 largely eliminates caching: tensor computations are driven by compiler- and runtime-managed data streams flowing between computing elements and on/off-chip memories, yielding greater efficiency and scalability than cache-based systems. Another key principle was selecting the right operational granularity for scheduling, issuing, and execution across compiler, firmware, and hardware. Recognizing commonalities in AI workloads, we chose the tensor as the fundamental data element. M100 demonstrates general AI computing capability across diverse inference applications, including UniAD (for AD) and LLaMA (for LLMs). Benchmarks show M100 outperforms GPGPU architectures in AD applications with higher utilization, representing a promising direction for future general AI computing.

cs.LG↗

arXiv2026

Discussion on the equivalence of two relativistic point-particle Lagrangians

In 2021, Lei et al. claimed the equivalence between the two Lagrangians $\mathcal{L}_1 =-mc\sqrt{-g_{μν}{\dot{x}}^μ{\dot{x}}^ν}-V$ and $\mathcal{L}_2 = \frac{1}{2}mg_{μν} {\dot{x}}^μ{\dot{x}}^ν-V$ for describing particle dynamics in combined gravitational and matter fields. In the present work, we rigorously demonstrate that their equivalence depends critically on the external potential V. Both Lagrangians yield identical Hamiltonians that strictly satisfy the mass shell constraint, and are therefore equivalent when V vanishes or corresponds to an electromagnetic potential. However, they are generally not equivalent for generic external potentials excluding the electromagnetic ones. This discrepancy arises because L1 and L2 correspond to different Hamiltonian formulations. The Hamiltonian derived from L1 inherently enforces the mass shell constraint, whereas the Hamiltonian from L2 does not. When the Schwarzschild metric supplemented with an artificial mechanical potential is taken as a toy model, numerical investigations reveal that L1 leads to chaotic behavior, which signifies non-integrable dynamics. By contrast, L2 can be shown analytically to produce integrable dynamics free of chaos. In many scenarios, L1 is strongly recommended due to its theoretical superiority and universality. L2 is generally suitable for classical approximate problems involving low energy and weak gravity. Nevertheless, it is the preferred choice for strong field problems concerning the dynamics of charged (or neutral) particles near black holes with (or without) external electromagnetic fields, owing to its mathematical simplicity and computational efficiency. Moreover, it can still satisfy the mass shell constraint when an additional constraint is imposed on its corresponding Hamiltonian.

gr-qc↗

arXiv2026

ST-BiBench: Benchmarking Multi-Stream Multimodal Coordination in Bimanual Embodied Tasks for MLLMs

Multimodal Large Language Models (MLLMs) have significantly advanced the landscape of embodied AI, yet transitioning to synchronized bimanual coordination introduces formidable challenges in multi-stream multimodal integration. We introduce ST-BiBench, a comprehensive multi-tier framework for evaluating spatio-temporal multimodal coordination. Our approach centers on Strategic Coordination Planning, assessing high-level cross-modal reasoning over multiple action and perception streams. To investigate the "proximity paradox"-where semantically coherent plans fail to align with spatially grounded visual inputs-we incorporate Foundational Spatial Grounding to verify workspace awareness and arm-selection logic. Furthermore, we probe model frontiers through Fine-Grained Action Control, investigating whether MLLMs can directly synthesize high-dimensional continuous action modalities (16-Dim) from complex multimodal metadata. Evaluating 30+ state-of-the-art MLLMs, we uncover a persistent and pervasive "coordination paradox"-a significant gap between high-level strategic reasoning and fine-grained physical execution. Results reveal that while frontier MLLMs excel at logic-driven strategy, they frequently suffer from perception-logic disconnection and multi-stream interference during multimodal fusion. ST-BiBench provides a platform for identifying critical bottlenecks in multi-stream multimodal fusion and cross-modal alignment for complex embodied tasks.

cs.RO↗

arXiv2026

KEditVis: A Visual Analytics System for Knowledge Editing of Large Language Models

Large Language Models (LLMs) demonstrate exceptional capabilities in factual question answering, yet they sometimes provide incorrect responses. To address this issue, knowledge editing techniques have emerged as effective methods for correcting factual information in LLMs. However, typical knowledge editing workflows struggle with identifying the optimal set of model layers for editing and rely on summary indicators that provide insufficient guidance. This lack of transparency hinders effective comparison and identification of optimal editing strategies. In this paper, we present KEditVis, a novel visual analytics system designed to assist users in gaining a deeper understanding of knowledge editing through interactive visualizations, improving editing outcomes, and discovering valuable insights for the future development of knowledge editing algorithms. With KEditVis, users can select appropriate layers as the editing target, explore the reasons behind ineffective edits, and perform more targeted and effective edits. Our evaluation, including usage scenarios, expert interviews, and a user study, validates the effectiveness and usability of the system.

cs.HC↗

arXiv2026

PRISM: Dynamic Primitive-Based Forecasting for Large-Scale GPU Cluster Workloads

Accurately forecasting GPU workloads is essential for AI infrastructure, enabling efficient scheduling, resource allocation, and power management. Modern workloads are highly volatile, multiple periodicity, and heterogeneous, making them challenging for traditional predictors. We propose PRISM, a primitive-based compositional forecasting framework combining dictionary-driven temporal decomposition with adaptive spectral refinement. This dual representation extracts stable, interpretable workload signatures across diverse GPU jobs. Evaluated on large-scale production traces, PRISM achieves state-of-the-art results. It significantly reduces burst-phase errors, providing a robust, architecture-aware foundation for dynamic resource management in GPU-powered AI platforms.

cs.DC↗

arXiv2026

GPUMDkit: A User-Friendly Toolkit for GPUMD and NEP

Machine-learned interatomic potentials have revolutionized molecular dynamics simulations by providing quantum-mechanical accuracy at empirical-potential speeds. The graphics processing unit molecular dynamics (GPUMD) package, featuring the highly efficient neuroevolution potential (NEP) framework, has emerged as a powerful tool in this domain. However, the complexity of force field development, active learning, and trajectory post-processing often requires extensive manual scripting, imposing a steep learning curve on new users. To address this, we present GPUMDkit, a comprehensive and user-friendly toolkit that streamlines the entire simulation workflow for GPUMD and NEP. GPUMDkit integrates a suite of essential functionalities, including format conversion, structure sampling, property calculation, and data visualization, accessible through both interactive and command-line interfaces. Its modular, extensible architecture ensures accessibility for users of all experience levels while allowing seamless integration of new features. By automating complex tasks and enhancing productivity, GPUMDkit substantially lowers the barrier to using GPUMD and NEP programs. This article describes the program architecture and demonstrates its capabilities through practical applications.

cond-mat.mtrl-sci↗

arXiv2026

Third type of spacetime with the coexistence of integrability and non-integrability

The integrability or non-integrability of a spacetime usually refers to whether the motion of massive or massless particles in the spacetime is integrable or not. The standard black hole spacetimes such as the Schwarzschild and Kerr metrics are always integrable for both timelike and null geodesics. They belong to a first type of spacetime. However, the Melvin type spacetimes as a second type of spacetime are non-integrable, regardless of whether they are for massive or massless particle motion. In this paper, we discover the possibility of a third type of spacetime with non-integrable dynamics of timelike geodesics and integrable dynamics of null geodesics. In fact, conformal transformations may transform type one solutions into type three. This is due to the conformal factors preventing the separation of variables from the Hamilton-Jacobi equation and leading to the absence of a fourth constant of motion for the massive particle dynamics. Nevertheless, the massless particle motion still remains integrable in these metrics for any conformal factors because the conformal factors have no effect on the null geodesics whatsoever. The conformal Kerr metric is an example of the third type of spacetime. In addition to the conformal transformation method, other paths may yield the third type of spacetime. The Kerr-Bertotti-Robinson black hole metric and the accelerating Schwarzschild spacetime are two examples of non-conformal solutions that are also of type three.

gr-qc↗

arXiv2026

Spatial4D-Bench: A Versatile 4D Spatial Intelligence Benchmark

4D spatial intelligence involves perceiving and processing how objects move or change over time. Humans naturally possess 4D spatial intelligence, supporting a broad spectrum of spatial reasoning abilities. To what extent can Multimodal Large Language Models (MLLMs) achieve human-level 4D spatial intelligence? In this work, we present Spatial4D-Bench, a versatile 4D spatial intelligence benchmark designed to comprehensively assess the 4D spatial reasoning abilities of MLLMs. Unlike existing spatial intelligence benchmarks that are often small-scale or limited in diversity, Spatial4D-Bench provides a large-scale, multi-task evaluation benchmark consisting of ~40,000 question-answer pairs covering 18 well-defined tasks. We systematically organize these tasks into six cognitive categories: object understanding, scene understanding, spatial relationship understanding, spatiotemporal relationship understanding, spatial reasoning and spatiotemporal reasoning. Spatial4D-Bench thereby offers a structured and comprehensive benchmark for evaluating the spatial cognition abilities of MLLMs, covering a broad spectrum of tasks that parallel the versatility of human spatial intelligence. We benchmark various state-of-the-art open-source and proprietary MLLMs on Spatial4D-Bench and reveal their substantial limitations in a wide variety of 4D spatial reasoning aspects, such as route plan, action recognition, and physical plausibility reasoning. We hope that the findings provided in this work offer valuable insights to the community and that our benchmark can facilitate the development of more capable MLLMs toward human-level 4D spatial intelligence. More resources can be found on our project page.

cs.CV↗

arXiv2026

SSR: Pushing the Limit of Spatial Intelligence with Structured Scene Reasoning

While Multimodal Large Language Models (MLLMs) excel in semantic tasks, they frequently lack the "spatial sense" essential for sophisticated geometric reasoning. Current models typically suffer from exorbitant modality-alignment costs and deficiency in fine-grained structural modeling precision.We introduce SSR, a framework designed for Structured Scene Reasoning that seamlessly integrates 2D and 3D representations via a lightweight alignment mechanism. To minimize training overhead, our framework anchors 3D geometric features to the large language model's pre-aligned 2D visual semantics through cross-modal addition and token interleaving, effectively obviating the necessity for large-scale alignment pre-training. To underpin complex spatial reasoning, we propose a novel scene graph generation pipeline that represents global layouts as a chain of independent local triplets defined by relative coordinates. This is complemented by an incremental generation algorithm, enabling the model to construct "language-model-friendly" structural scaffolds for complex environments. Furthermore, we extend these capabilities to global-scale 3D global grounding task, achieving absolute metric precision across heterogeneous data sources. At a 7B parameter scale, SSR achieves state-of-the-art performance on multiple spatial intelligence benchmarks, notably scoring 73.9 on VSI-Bench. Our approach significantly outperforms much larger models, demonstrating that efficient feature alignment and structured scene reasoning are the cornerstones of authentic spatial intelligence.

cs.CV↗

arXiv2026

Calculations of Di-Hadron Production via Two-Photon Processes in Relativistic Heavy-Ion Collisions

Two-photon processes in relativistic heavy-ion collisions have emerged as a critical probe of quantum electrodynamics in ultra-intense electromagnetic fields, with recent focus extending beyond dileptons to hadronic final states. At present, quantitative studies of di-hadron production via two-photon interactions remain scarce. In this work, we employ the Equivalent Photon Approximation and the two-photon fusion measurements from $e^{+}e^{-}$ collisions to obtain differential cross-section predictions for $π^{+}π^{-}$, $K^{+}K^{-}$, and $p\bar{p}$ pairs produced in ultra-peripheral $\mathrm{Au{+}Au}$ collisions at $\sqrt{s_{NN}} = 200\,\text{GeV}$ within the STAR acceptance, as well as in $\mathrm{Pb{+}Pb}$ collisions at $\sqrt{s_{NN}} = 5.36\,\text{TeV}$ within typical LHC acceptance. The calculations deliver the unified baseline for light-meson and baryon pairs in this environment, supplying benchmarks for upcoming STAR and LHC measurements and guiding future systematic investigations of hadronic two-photon processes at RHIC and LHC facilities.

hep-ph↗