arXiv subjects

Xin Liu

Publications and source records attributed to Xin Liu.

At least 199 records · Page 11Linked to original sources

arXiv2026

Edge-Assisted Multi-Robot Visual-Inertial SLAM with Efficient Communication

The integration of cloud computing and edge computing is an effective way to achieve global consistent and real-time multi-robot Simultaneous Localization and Mapping (SLAM). Cloud computing effectively solves the problem of limited computing, communication and storage capacity of terminal equipment. However, limited bandwidth and extremely long communication links between terminal devices and the cloud result in serious performance degradation of multi-robot SLAM systems. To reduce the computational cost of feature tracking and improve the real-time performance of the robot, a lightweight SLAM method of optical flow tracking based on pyramid IMU prediction is proposed. On this basis, a centralized multi-robot SLAM system based on a robot-edge-cloud layered architecture is proposed to realize real-time collaborative SLAM. It avoids the problems of limited on-board computing resources and low execution efficiency of single robot. In this framework, only the feature points and keyframe descriptors are transmitted and lossless encoding and compression are carried out to realize real-time remote information transmission with limited bandwidth resources. This design reduces the actual bandwidth occupied in the process of data transmission, and does not cause the loss of SLAM accuracy caused by data compression. Through experimental verification on the EuRoC dataset, compared with the current most advanced local feature compression method, our method can achieve lower data volume feature transmission, and compared with the current advanced centralized multi-robot SLAM scheme, it can achieve the same or better positioning accuracy under low computational load.

cs.RO↗

arXiv2026

Property-driven Protein Inverse Folding With Multi-Objective Preference Alignment

Protein sequence design must balance designability, defined as the ability to recover a target backbone, with multiple, often competing, developability properties such as solubility, thermostability, and expression. Existing approaches address these properties through post hoc mutation, inference-time biasing, or retraining on property-specific subsets, yet they are target dependent and demand substantial domain expertise or careful hyperparameter tuning. In this paper, we introduce ProtAlign, a multi-objective preference alignment framework that fine-tunes pretrained inverse folding models to satisfy diverse developability objectives while preserving structural fidelity. ProtAlign employs a semi-online Direct Preference Optimization strategy with a flexible preference margin to mitigate conflicts among competing objectives and constructs preference pairs using in silico property predictors. Applied to the widely used ProteinMPNN backbone, the resulting model MoMPNN enhances developability without compromising designability across tasks including sequence design for CATH 4.3 crystal structures, de novo generated backbones, and real-world binder design scenarios, making it an appealing framework for practical protein sequence design.

cs.LG↗

arXiv2026

SynthWorlds: Controlled Parallel Worlds for Disentangling Reasoning and Knowledge in Language Models

Evaluating the reasoning ability of language models (LMs) is complicated by their extensive parametric world knowledge, where benchmark performance often reflects factual recall rather than genuine reasoning. Existing datasets and approaches (e.g., temporal filtering, paraphrasing, adversarial substitution) cannot cleanly separate the two. We present SynthWorlds, a framework that disentangles task reasoning complexity from factual knowledge. In SynthWorlds, we construct parallel corpora representing two worlds with identical interconnected structure: a real-mapped world, where models may exploit parametric knowledge, and a synthetic-mapped world, where such knowledge is meaningless. On top of these corpora, we design two mirrored tasks as case studies: multi-hop question answering and page navigation, which maintain equal reasoning difficulty across worlds. Experiments in parametric-only (e.g., closed-book QA) and knowledge-augmented (e.g., retrieval-augmented) LM settings reveal a persistent knowledge advantage gap, defined as the performance boost models gain from memorized parametric world knowledge. Knowledge acquisition and integration mechanisms reduce but do not eliminate this gap, highlighting opportunities for system improvements. Fully automatic and scalable, SynthWorlds provides a controlled environment for evaluating LMs in ways that were previously challenging, enabling precise and testable comparisons of reasoning and memorization.

cs.CL↗

arXiv2026

Diff-Muscle: Efficient Learning for Musculoskeletal Robotic Table Tennis

Musculoskeletal robots provide superior advantages in flexibility and dexterity, positioning them as a promising frontier towards embodied intelligence. However, current research is largely confined to relative simple tasks, restricting the exploration of their full potential in multi-segment coordination. Furthermore, efficient learning remains a challenge, primarily due to the high-dimensional action space and inherent overactuated structures. To address these challenges, we propose Diff-Muscle, a musculoskeletal robot control algorithm that leverages differential flatness to reformulate policy learning from the redundant muscle-activation space into a significantly lower-dimensional joint space. Furthermore, we utilize the highly dynamic robotic table tennis task to evaluate our algorithm. Specifically, we propose a hierarchical reinforcement learning framework that integrates a Kinematics-based Muscle Actuation Controller (K-MAC) with high-level trajectory planning, enabling a musculoskeletal robot to perform dexterous and precise rallies. Experimental results demonstrate that Diff-Muscle significantly outperforms state-of-the-art baselines in success rates while maintaining minimal muscle activation. Notably, the proposed framework successfully enables the musculoskeletal robots to achieve continuous rallies in a challenging dual-robot setting.

cs.RO↗

arXiv2026

How Well Do Multimodal Models Reason on ECG Signals?

While multimodal large language models offer a promising solution to the "black box" nature of health AI by generating interpretable reasoning traces, verifying the validity of these traces remains a critical challenge. Existing evaluation methods are either unscalable, relying on manual clinician review, or superficial, utilizing proxy metrics (e.g. QA) that fail to capture the semantic correctness of clinical logic. In this work, we introduce a reproducible framework for evaluating reasoning in ECG signals. We propose decomposing reasoning into two distinct, components: (i) Perception, the accurate identification of patterns within the raw signal, and (ii) Deduction, the logical application of domain knowledge to those patterns. To evaluate Perception, we employ an agentic framework that generates code to empirically verify the temporal structures described in the reasoning trace. To evaluate Deduction, we measure the alignment of the model's logic against a structured database of established clinical criteria in a retrieval-based approach. This dual-verification method enables the scalable assessment of "true" reasoning capabilities.

cs.AI↗

arXiv2026

Disentangled Textual Priors for Diffusion-based Image Super-Resolution

Image Super-Resolution (SR) aims to reconstruct high-resolution images from degraded low-resolution inputs. While diffusion-based SR methods offer powerful generative capabilities, their performance heavily depends on how semantic priors are structured and integrated into the generation process. Existing approaches often rely on entangled or coarse-grained priors that mix global layout with local details, or conflate structural and textural cues, thereby limiting semantic controllability and interpretability. In this work, we propose DTPSR, a novel diffusion-based SR framework that introduces disentangled textual priors along two complementary dimensions: spatial hierarchy (global vs. local) and frequency semantics (low- vs. high-frequency). By explicitly separating these priors, DTPSR enables the model to simultaneously capture scene-level structure and object-specific details with frequency-aware semantic guidance. The corresponding embeddings are injected via specialized cross-attention modules, forming a progressive generation pipeline that reflects the semantic granularity of visual content, from global layout to fine-grained textures. To support this paradigm, we construct DisText-SR, a large-scale dataset containing approximately 95,000 image-text pairs with carefully disentangled global, low-frequency, and high-frequency descriptions. To further enhance controllability and consistency, we adopt a multi-branch classifier-free guidance strategy with frequency-aware negative prompts to suppress hallucinations and semantic drift. Extensive experiments on synthetic and real-world benchmarks show that DTPSR achieves high perceptual quality, competitive fidelity, and strong generalization across diverse degradation scenarios.

cs.CV↗

arXiv2026

SpecFuse: Ensembling Large Language Models via Next-Segment Prediction

Ensembles of generative large language models (LLMs) are a promising way to compensate for individual model limitations, integrating the strengths of different LLMs. Existing LLM ensemble methods, however, face limitations such as first-token delay and challenges in long-range semantic collaboration between models, Moreover, they typically assume equal voting weights for all models during ensemble, ignoring task-specific performance differences among models. In this work, we propose SpecEM, a training-free, plug-and-play LLM ensemble framework that dynamically adjusts each model's model contribution in real time based on task performance. Inspired by speculative decoding, SpecEM iteratively performs drafting and verification, allowing models to collaborate semantically at the segment level for integrated output. Furthermore, we introduce an online feedback mechanism with multiplicative weight updates, where each model's voting weight is adjusted on-the-fly according to how often it outperforms others during verification stage, ensuring that stronger models exert greater influence during ensembling. Experimental results on five LLM families (ranging from 7B to 72B parameters) and six benchmark datasets, spanning open-domain instruction following, reasoning, commonsense, demonstrate consistent performance improvements compared to state-of-the-art LLM ensemble methods. Our code is available at https://github.com/lvbotenbest/SpecEM.

cs.CL↗

arXiv2026

Fingerprint Recognition of Partial Discharge Signals in Deep Learning Enhanced Rydberg Atomic Sensors

Partial discharge originates from microscopic insulation imperfections in high-voltage apparatus and is widely considered a critical marker of incipient deterioration. Conventional partial discharge detection methods are typically constrained by limited bandwidth and often rely on predefined feature extraction, which impedes reliable recognition of broadband transient signals. In this work, we employ a Rydberg atomic sensor to directly capture time-domain responses of partial discharge emissions and construct distinctive spectral fingerprints for different types. A 1D ResNet deep learning model is then applied to recognize these fingerprints from time-domain signals without manual feature engineering. Under increased source-antenna distances, where spectral features are significantly attenuated, the model attains a recognition accuracy of approximately 94\% across four partial discharge categories, demonstrating robustness to attenuation and noise. We further validate the approach in a simulated early-warning scenario, where partial discharge signals mixed with noise are analyzed and the model successfully generates predictive alarms. These results underscore the potential of integrating Rydberg-based broadband sensing with data-driven analysis for non-invasive, high-sensitivity diagnostics of electrical insulation systems.

quant-ph↗

arXiv2026

Toward Trustworthy Earthquake Catalogs in the Era of Automated Detection: A Probabilistic Framework for Robust Earthquake Location

The rapid proliferation of deep-learning-based detection and association methods has greatly expanded automatically generated earthquake catalogs, but has also introduced false detections, mis-associated arrivals, and poorly constrained events, making rigorous uncertainty quantification essential. We present a fully probabilistic earthquake location framework that jointly infers hypocenters, origin times, phase-dependent noise scales, and contamination levels within a unified Bayesian formulation. Robustness is achieved through a two-level hierarchical strategy: arrival-time residuals are modeled using a Student-$t$ scale-mixture to accommodate heavy-tailed noise, while an explicit two-component contamination model probabilistically classifies each phase pick as an inlier or outlier, with phase-specific contamination rates inferred from the data. This formulation avoids heuristic data rejection and manual thresholding. Posterior sampling is accelerated using a neural-network travel-time surrogate, enabling scalable inference for large catalogs. Synthetic tests demonstrate well-calibrated posterior uncertainties, and application to the 2022 Luding $M_s$~6.8 aftershock sequence shows that uncertainty-based screening reduces the catalog from 10{,}590 to 6{,}562 events without loss of recall. This framework provides a principled pathway toward statistically trustworthy earthquake catalogs in the era of automated seismic monitoring.

physics.geo-ph↗

arXiv2026

CUDA Agent: Large-Scale Agentic RL for High-Performance CUDA Kernel Generation

GPU kernel optimization is fundamental to modern deep learning but remains a highly specialized task requiring deep hardware expertise. Despite strong performance in general programming, large language models (LLMs) remain uncompetitive with compiler-based systems such as torch.compile for CUDA kernel generation. Existing CUDA code generation approaches either rely on training-free refinement or fine-tune models within fixed multi-turn execution-feedback loops, but both paradigms fail to fundamentally improve the model's intrinsic CUDA optimization ability, resulting in limited performance gains. We present CUDA Agent, a large-scale agentic reinforcement learning system that develops CUDA kernel expertise through three components: a scalable data synthesis pipeline, a skill-augmented CUDA development environment with automated verification and profiling to provide reliable reward signals, and reinforcement learning algorithmic techniques enabling stable training. CUDA Agent achieves state-of-the-art results on KernelBench, delivering 100\%, 100\%, and 92\% faster rate over torch.compile on KernelBench Level-1, Level-2, and Level-3 splits, outperforming the strongest proprietary models such as Claude Opus 4.5 and Gemini 3 Pro by about 40\% on the hardest Level-3 setting.

cs.LG↗

arXiv2026

Zero-Waiting Load Balancing with Heterogeneous Servers in Heavy Traffic

We study the steady-state delay performance of load balancing in large-scale systems with heterogeneous servers in the heavy-traffic regimes. The system consists of $N$ servers, each with a local buffer of size $b-1$, serving jobs in the first-in-first-out (FIFO) order. Jobs arrive according to a Poisson process with rate $λN$, where $λ= 1 - N^{-α}$ for any $α\in (0,1)$. Service times are assumed to be exponentially distributed with fully heterogeneous rates, where the service rate of each server can differ and may scale with the system size $N$. We study a queue length aware and service rate aware load balancing policy, Join-the-Fastest-Shortest-Queue (JFSQ), and demonstrate that it achieves asymptotic zero waiting time and probability under the heavy traffic regimes, including both the Sub-Halfin-Whitt ($α\in (0,0.5)$) and Super-Halfin-Whitt ($α\in [0.5,1)$) regimes. The performance bounds of waiting time and probability explicitly capture the convergence rate w.r.t. the system size $N$ and show the negative effect of server heterogeneity. Our analysis builds on the general framework of Stein's method with iterative state-space peeling, where we design a sequence of Lyapunov functions to analyze the high-dimensional heterogeneous system without assuming exchangeability and monotonicity. Our analysis shows that JFSQ efficiently utilizes servers with higher capacities, and the steady-state system can be coupled with a single-server queue via Stein's method. To the best of our knowledge, this is the first work to establish delay performance bounds of a load-balancing system with size $N$ and fully heterogeneous servers in heavy traffic.

math.PR↗

arXiv2026

Local well-posedness of strong solutions to the non-isentropic compressible primitive equations with vertical diffusion

Due to the absence of dynamical equation in the vertical momentum component of the primitive equations (PEs) of atmospheric dynamics, the vertical component of the velocity can be recovered only from the information on the other physical quantities, while utilizing the hydrostatic balance. This causes one spatial derivative loss while leads to a stronger nonlinearity, comparing to the classic compressible Navier-Stokes equations. As a result, the mathematical analysis on the compressible primitive equations is mathematically more challenging than that on the compressible Navier-Stokes equations. In this paper, we consider the initial-boundary value problem to the non-isentropic compressible primitive equations with only vertical diffusion for the temperature, but without gravity. Local existence and uniqueness as well as the continuous dependence on the initial data of strong solutions are established for any suitably regular initial data. The initial velocity and pressure are assumed here to be of one order derivative higher regularity than that of the initial density.

math.AP↗

arXiv2026

Full-Field Metasurface Characterization with Polarization Sensitive Coherent Modulation Imaging

Characterizing the intensity, phase, and polarization of engineered light is fundamental to understanding and applying metasurfaces. However, existing characterization frameworks are hindered by several limitations, most notably their inability to account for the polarization of the field. Here, we report polarization sensitive coherent modulation imaging (PS-CMI), a light-weight but robust, high-resolution platform for the full-field characterization of metasurface-modulated light. By supplementing the orthogonal x- and y- complex amplitude components with an additional 45°-component, this approach calculates the retardance between two orthogonal polarization components while eliminating phase offsets, thereby enabling the subsequent recovery of the complete polarization state. We demonstrate the versatility of our method by characterizing light fields produced by a United States Air Force (USAF) target, two kinds of complex polarization field, and a metalens. This compact solution addresses a critical gap in metasurface metrology and is broadly applicable to other fields requiring the mapping of complex, polarized light distributions.

physics.optics↗

arXiv2026

A Scalable Framework for Evaluating Health Language Models

Large language models (LLMs) have emerged as powerful tools for analyzing complex datasets. Recent studies demonstrate their potential to generate useful, personalized responses when provided with patient-specific health information that encompasses lifestyle, biomarkers, and context. As LLM-driven health applications are increasingly adopted, rigorous and efficient one-sided evaluation methodologies are crucial to ensure response quality across multiple dimensions, including accuracy, personalization and safety. Current evaluation practices for open-ended text responses heavily rely on human experts. This approach introduces human factors and is often cost-prohibitive, labor-intensive, and hinders scalability, especially in complex domains like healthcare where response assessment necessitates domain expertise and considers multifaceted patient data. In this work, we introduce Adaptive Precise Boolean rubrics: an evaluation framework that streamlines human and automated evaluation of open-ended questions by identifying gaps in model responses using a minimal set of targeted rubrics questions. Our approach is based on recent work in more general evaluation settings that contrasts a smaller set of complex evaluation targets with a larger set of more precise, granular targets answerable with simple boolean responses. We validate this approach in metabolic health, a domain encompassing diabetes, cardiovascular disease, and obesity. Our results demonstrate that Adaptive Precise Boolean rubrics yield higher inter-rater agreement among expert and non-expert human evaluators, and in automated assessments, compared to traditional Likert scales, while requiring approximately half the evaluation time of Likert-based methods. This enhanced efficiency, particularly in automated evaluation and non-expert contributions, paves the way for more extensive and cost-effective evaluation of LLMs in health.

cs.AI↗

arXiv2026

Dual-Granularity Contrastive Reward via Generated Episodic Guidance for Efficient Embodied RL

Designing suitable rewards poses a significant challenge in reinforcement learning (RL), especially for embodied manipulation. Trajectory success rewards are suitable for human judges or model fitting, but the sparsity severely limits RL sample efficiency. While recent methods have effectively improved RL via dense rewards, they rely heavily on high-quality human-annotated data or abundant expert supervision. To tackle these issues, this paper proposes Dual-granularity contrastive reward via generated Episodic Guidance (DEG), a novel framework to seek sample-efficient dense rewards without requiring human annotations or extensive supervision. Leveraging the prior knowledge of large video generation models, DEG only needs a small number of expert videos for domain adaptation to generate dedicated task guidance for each RL episode. Then, the proposed dual-granularity reward that balances coarse-grained exploration and fine-grained matching, will guide the agent to efficiently approximate the generated guidance video sequentially in the contrastive self-supervised latent space, and finally complete the target task. Extensive experiments on 18 diverse tasks across both simulation and real-world settings show that DEG can not only serve as an efficient exploration stimulus to help the agent quickly discover sparse success rewards, but also guide effective RL and stable policy convergence independently.

cs.LG↗

arXiv2026

Fast convolution solvers using moment-matching

We propose two easy-to-implement fast algorithms based on moment-matching to compute the nonlocal potential $φ(\textbf{x})=(U\ast ρ)(\textbf{x})$ on bounded domain, where the kernel $U$ is singular at the origin and the density $ρ$ is a fast-decaying smooth function. Each method requires merely minor modifications to commonly-used existing methods, i.e., the sine spectral/Fourier quadrature method, and achieves a much better convergence rate. The key lies in the introduction of a smooth auxiliary function $ρ_1$ whose moments match those of the density up to an integer order $m$. Specifically, $ρ_1$ is constructed using Gaussian function in an explicit way and the associated potential can be calculated analytically. The moments of residual density vanish up to order $m$, and the corresponding residual potential $U \ast (ρ-ρ_1)$ decays much faster than the original potential $φ$ at the far field. As for the residual potential evaluation, for classical kernels (e.g., the Coulomb kernel), we solve a differential/pseudo-differential equation on a rectangular domain with homogeneous Dirichlet boundary conditions via sine pseudospectral method, and achieve an arbitrary high convergence rate. While, for general kernels, the regularity of Fourier integrand increase by $m$ thanks to the moments-vanishing property, therefore, the standard trapezoidal rule/midpoint quadrature also converges much faster. To gain a better numerical performance, we utilize the domain expansion technique to obtain better accuracy, and improve the efficiency by simplifying the quadrature into one discrete convolution and applying Fast Fourier Transform (FFT) to a double-sized vector. Rigorous error estimates and extensive numerical investigations showcase the accuracy and efficiency for different nonlocal potentials.

math.NA↗

arXiv2026

ABot-N0: Technical Report on the VLA Foundation Model for Versatile Embodied Navigation

Embodied navigation has long been fragmented by task-specific architectures. We introduce ABot-N0, a unified Vision-Language-Action (VLA) foundation model that achieves a ``Grand Unification'' across 5 core tasks: Point-Goal, Object-Goal, Instruction-Following, POI-Goal, and Person-Following. ABot-N0 utilizes a hierarchical ``Brain-Action'' architecture, pairing an LLM-based Cognitive Brain for semantic reasoning with a Flow Matching-based Action Expert for precise, continuous trajectory generation. To support large-scale learning, we developed the ABot-N0 Data Engine, curating 16.9M expert trajectories and 5.0M reasoning samples across 7,802 high-fidelity 3D scenes (10.7 $\text{km}^2$). ABot-N0 achieves new SOTA performance across 7 benchmarks, significantly outperforming specialized models. Furthermore, our Agentic Navigation System integrates a planner with hierarchical topological memory, enabling robust, long-horizon missions in dynamic real-world environments.

cs.RO↗

arXiv2026

Theoretical Modeling of Large Language Model Self-Improvement Training Dynamics Through Solver-Verifier Gap

Self-improvement is a significant techniques within the realm of large language model (LLM), aiming to enhance the LLM performance without relying on external data. Despite its significance, generally how LLM performances evolve during the self-improvement process remains underexplored. In this paper, we theoretically model the training dynamics of self-improvement via the concept of solver-verifier gap. This is inspired by the conjecture that the performance enhancement of self-improvement stems from the gap between LLM's solver capability and verifier capability. Based on the theoretical framework, we further show how to model the entire training trajectory. This framework allows quantifying the capability limit of self-improvement by fitting the theoretical model to the experiment results. We validate the effectiveness of the theoretical framework on various LLMs and datasets. Beyond self-improvement, we extend our analysis to investigate how external data influences these dynamics within the framework. Notably, we find that under limited external data regimes, such external data can be utilized at any stage without significantly affecting final performances, which accords with the empirical observations.

cs.LG↗