arXiv subjects

Lu Zhao

Publications and source records attributed to Lu Zhao.

At least 19 recordsLinked to original sources

arXiv2026

PACE: Adaptive Budget Allocation for Time-Efficient Embodied Planning

Reasoning-enhanced large language models have achieved remarkable improvements in planning tasks, yet their deployment in embodied systems remains impractical due to prohibitive inference delays-often exceeding minutes per planning instance. The fundamental bottleneck stems from the serial nature of existing paradigms: models must complete all reasoning before any action execution, leaving execution time windows entirely unexploited. We introduce PACE (Planning with Adaptive Cognitive Effort), a framework that enables interleaved reasoning and execution through two key innovations: an Interleaved Think-Act architecture that pipelines cognitive processing with action execution, and a Dynamic Budget Allocator that adapts reasoning token budgets to available execution time windows. On the Robotouille benchmark using Qwen3-8B-AWQ, PACE achieves a 10% success rate-representing a 67% improvement over the ReAct+Think baseline-while delivering 6.9 times acceleration in thinking time compared to unconstrained reasoning. The framework hides 66.8% of thinking time within execution windows, demonstrating that strategic cognitive effort allocation can simultaneously improve both planning quality and time efficiency. These results provide evidence that time-aware architectural innovations enable reasoning models to operate in latency-sensitive embodied domains where they were previously impractical.

cs.RO↗

arXiv2026

Direct sampling methods for inverse medium scattering problems of elastic waves

This paper concerns the inverse elastic scattering problem of determining an unknown penetrable obstacle from far-field data. Using the Helmholtz decomposition, the coupled boundary value problem is reformulated as a coupled scalar Helmholtz system. We prove the uniqueness of the associated coupled Helmholtz boundary value problem and of the corresponding boundary integral equation system, which is then discretized by a Nystr\"om method for efficient numerical computation. Leveraging the relation between the compressional and shear far-field patterns of the Navier system and those of the coupled Helmholtz system, we employ three indicators to reconstruct the location and shape of the obstacle. Furthermore, we analyze the decay properties of these indicators and establish corresponding stability estimates. Numerical experiments are presented to illustrate the effectiveness and robustness of the proposed method, even for limited-aperture data.

math.NA↗

arXiv2026

A novel time-domain iterative method for a three-dimensional inverse acoustic obstacle scattering problem

This paper concerns the three-dimensional forward and inverse acoustic obstacle scattering problem in the time domain. For the forward problem, a retarded potential formulation discretized by convolution quadrature and Galerkin methods is introduced. By introducing the retarded boundary integral defined on a homothetic surface, we propose a novel time-domain convolution quadrature based iterative method to reconstruct both the shape and location of a rigid obstacle. The retarded integral in the time domain is reformulated into a system of integrals in the s-domain. The resulting s-domain integrals are very fast to compute, as they only involve non-singular integrals over the homothetic surfaces. Moreover, the Fr\'echet derivative with respect to the boundary can be derived straightforwardly. We also prove that the scattered field generated by the homothetic surface converges to the exact field in the time domain. To improve the stability of the inversion algorithm, an incremental truncation technique is proposed, and numerical experiments confirm the effectiveness and robustness of our method.

math.NA↗

arXiv2026

A highly efficient iterative approach for inverse acoustic obstacle scattering problems in three dimensions

This paper concerns a three-dimensional inverse acoustic obstacle scattering problem from scattered field or phased/phaseless far-field data. Based on the boundary integral defined on a homothetic surface, we propose a highly efficient iterative approach for obstacle reconstruction that completely avoids dealing with any singularity. Here, the injectivity and dense-range property of the Fr\'echet derivative have been proved to ensure the solvability of the linearized equivalent data equation. We also prove that the scattered field generated by the homothetic surface can arbitrarily approximate the exact one. Numerical experiments are presented to verify the superiority and robustness of the proposed approach.

math.NA↗

arXiv2026

NTIRE 2026 Challenge on Single Image Reflection Removal in the Wild: Datasets, Results, and Methods

In this paper, we review the NTIRE 2026 challenge on single-image reflection removal (SIRR) in the wild. SIRR is a fundamental task in image restoration. Despite progress in academic research, most methods are tested on synthetic images or limited real-world images, creating a gap in real-world applications. In this challenge, we provide participants with the OpenRR-5k dataset. This dataset requires participants to process real-world images covering a range of reflection scenarios and intensities, aiming to generate clean images without reflections. The challenge attracted more than 100 registrations, with eleven of them participating in the final testing phase. The top-ranked methods advanced the state-of-the-art reflection removal performance and earned unanimous recognition from five experts in the field. The proposed OpenRR-5k dataset is available at https://huggingface.co/datasets/qiuzhangTiTi/OpenRR-5k, and the homepage of this challenge is at https://github.com/caijie0620/OpenRR-5k.

cs.CV↗

arXiv2025

MemFine: Memory-Aware Fine-Grained Scheduling for MoE Training

The training of large-scale Mixture of Experts (MoE) models faces a critical memory bottleneck due to severe load imbalance caused by dynamic token routing. This imbalance leads to memory overflow on GPUs with limited capacity, constraining model scalability. Existing load balancing methods, which cap expert capacity, compromise model accuracy and fail on memory-constrained hardware. To address this, we propose MemFine, a memory-aware fine-grained scheduling framework for MoE training. MemFine decomposes the token distribution and expert computation into manageable chunks and employs a chunked recomputation strategy, dynamically optimized through a theoretical memory model to balance memory efficiency and throughput. Experiments demonstrate that MemFine reduces activation memory by 48.03% and improves throughput by 4.42% compared to full recomputation-based baselines, enabling stable large-scale MoE training on memory-limited GPUs.

cs.DC↗

arXiv2025

MoFa: A Unified Performance Modeling Framework for LLM Pretraining

The exponential growth in LLM scales, with parameters soaring from billions to trillions, has necessitated distributed pretraining across large clusters comprising thousands to tens of thousands of devices. While hybrid parallelization strategies enable such pretraining, the vast combinatorial strategy space introduces significant optimization challenges. Traditional manual tuning methods incur prohibitive trial-and-error costs, and existing performance modeling approaches exhibit critical limitations: they fail to comprehensively account for prevalent optimization features and ignore the substantial overhead imposed by essential fault tolerance mechanisms like checkpoint recovery in long-duration pretraining. To address these gaps, we propose MoFa, a novel pretraining performance modeling framework that unifies multi-dimensional optimization features and fault tolerance. MoFa incorporates an enhanced cost model to accurately capture the effects of key optimizations and integrates a fault tolerance model based on historical cluster reliability data. Besides, a MoFa-based tuning system is developed to explore optimal pretraining performance and potential bottlenecks in various scenarios. Extensive modeling evaluations demonstrate that MoFa can achieve high prediction accuracy across various scenarios. In addition, through comprehensive tuning experiments, our framework systematically reveals the key factors influencing pretraining performance under different configurations, which provides solid a priori guidance for LLM pretraining system design and deployment.

cs.DC↗

arXiv2025

PIMfused: Near-Bank DRAM-PIM with Fused-layer Dataflow for CNN Data Transfer Optimization

Near-bank Processing-in-Memory (PIM) architectures integrate processing cores (PIMcores) close to DRAM banks to mitigate the high cost of off-chip memory accesses. When accelerating convolutional neural network (CNN) on DRAM-PIM, performance is often constrained by cross-bank (or cross-PIMcore) data transfers, which are induced by the conventional layer-by-layer dataflow that enforces inter-bank (or inter-PIMcore) dependencies across successive CNN layers. To address this challenge, we propose PIMfused, a hardware-software co-design that enables fused-layer dataflow for end-to-end CNN execution in near-bank DRAM-PIM. By adopting fused-layer dataflow, PIMfused improves data reuse and, more importantly, breaks inter-bank data dependencies, thereby optimizing cross-bank data transfers without sacrificing bank-level parallelism. We study the impact of buffer sizes and PIMcore parallelism (1-bank vs. 4-bank) on PIMfused using end-to-end ResNet18. We present three key takeaways and show that with 4-bank PIMcores, PIMfused achieves overall PPA gains over a GDDR6-AiM-like baseline, cutting memory cycles to 30.6%, energy to 83.4%, and area to 76.5%.

cs.AR↗

arXiv2025

Disaggregated Prefill and Decoding Inference System for Large Language Model Serving on Multi-Vendor GPUs

LLM-based applications have been widely used in various industries, but with the increasing of models size, an efficient large language model (LLM) inference system is an urgent problem to be solved for service providers. Since the inference system is divided into two stage with different characteristics: Prefill and Decode, the two stage will interfere with each other during the inference process. Toward this end, a P-D disaggregated inference framework is proposed by some researchers. Current research is done on homogeneous GPUs, and lacks deployment solutions based on business scenarios. Compared with homogeneous GPUs, using heterogeneous GPUs to construct inference systems can better improve resource utilization and reduce costs. Even if GPUs from different vendors are used to build inference systems, on the basis of reducing costs, the resource utilization rate can be improved and the dependence on a single vendor can be reduced. Therefore, a P-D disaggreagetd inference system based on heterogeneous GPUs is designed, and the heterogeneous compatible transmission module in the system is designed to address heterogeneous GPU data compatibility issues. Then, a joint optimization algorithm of parallel strategy and instance number allocation is proposed to obtain the deployment solutions. Finally, the experimental results show that the P-D disaggregated inference system can well solve the hybrid inference problem of heterogeneous GPUs from different vendors, and the joint optimization algorithm can obtain the optimal deployment solution.

cs.DC↗

arXiv2025

DocIQ: A Benchmark Dataset and Feature Fusion Network for Document Image Quality Assessment

Document image quality assessment (DIQA) is an important component for various applications, including optical character recognition (OCR), document restoration, and the evaluation of document image processing systems. In this paper, we introduce a subjective DIQA dataset DIQA-5000. The DIQA-5000 dataset comprises 5,000 document images, generated by applying multiple document enhancement techniques to 500 real-world images with diverse distortions. Each enhanced image was rated by 15 subjects across three rating dimensions: overall quality, sharpness, and color fidelity. Furthermore, we propose a specialized no-reference DIQA model that exploits document layout features to maintain quality perception at reduced resolutions to lower computational cost. Recognizing that image quality is influenced by both low-level and high-level visual features, we designed a feature fusion module to extract and integrate multi-level features from document images. To generate multi-dimensional scores, our model employs independent quality heads for each dimension to predict score distributions, allowing it to learn distinct aspects of document image quality. Experimental results demonstrate that our method outperforms current state-of-the-art general-purpose IQA models on both DIQA-5000 and an additional document image dataset focused on OCR accuracy.

cs.CV↗

arXiv2025

Mechanistic Insight into BEOL Thermal Transport via Optical Metrology and Multiphysics Simulation

As integrated circuits continue to scale down and adopt three-dimensional (3D) stacking, thermal management in the back-end-of-line (BEOL) has emerged as a critical design constraint. In this study, we present a combined experimental and simulation framework to quantitatively characterize and mechanistically understand thermal transport in BEOL multilayers. Using the Square-Pulsed Source (SPS) method, a time-resolved optical metrology technique, we measure cross-plane thermal resistance and areal heat capacity in semiconductor chips at nanometer resolution. Two fabricated chip samples, polished to the M4 and M6 interconnection layers, are analyzed to extract thermal properties of distinct multilayer stacks. Results show that thermal resistance follows a series model, while areal heat capacity scales linearly with metal content. To uncover the underlying physical mechanisms, we perform finite element simulations using COMSOL Multiphysics, examining the influence of via connectivity and dielectric thermal conductivity on effective cross-plane heat transport. The simulations reveal that dielectric materials, due to their large volume fraction, are the primary limiting factor in BEOL thermal conduction, while the via structure plays a secondary but significant role. This combined experimental-simulation approach provides mechanistic insight into heat transport in advanced IC architectures and offers practical guidance for optimizing thermal pathways in future high-performance 3D-stacked devices.

cond-mat.mes-hall↗

arXiv2025

NTIRE 2025 Challenge on Image Super-Resolution (x4): Methods and Results

This paper presents the NTIRE 2025 image super-resolution ($\times$4) challenge, one of the associated competitions of the 10th NTIRE Workshop at CVPR 2025. The challenge aims to recover high-resolution (HR) images from low-resolution (LR) counterparts generated through bicubic downsampling with a $\times$4 scaling factor. The objective is to develop effective network designs or solutions that achieve state-of-the-art SR performance. To reflect the dual objectives of image SR research, the challenge includes two sub-tracks: (1) a restoration track, emphasizes pixel-wise accuracy and ranks submissions based on PSNR; (2) a perceptual track, focuses on visual realism and ranks results by a perceptual score. A total of 286 participants registered for the competition, with 25 teams submitting valid entries. This report summarizes the challenge design, datasets, evaluation protocol, the main results, and methods of each team. The challenge serves as a benchmark to advance the state of the art and foster progress in image SR.

cs.CV↗

arXiv2023

Bounding free energy difference with flow matching

This paper introduces a method for computing the Helmholtz free energy using the flow matching technique. Unlike previous work that utilized flow-based models for variational free energy calculations, this method provides bounds for free energy estimation based on targeted free energy perturbation, by performing calculations on samples from both ends of the mapping. We demonstrate applications of the present method by estimating the free energy of the classical Coulomb gas in a harmonic trap.

physics.comp-ph↗

arXiv2022

STN: Scalable Tensorizing Networks via Structure-Aware Training and Adaptive Compression

Deep neural networks (DNNs) have delivered a remarkable performance in many tasks of computer vision. However, over-parameterized representations of popular architectures dramatically increase their computational complexity and storage costs, and hinder their availability in edge devices with constrained resources. Regardless of many tensor decomposition (TD) methods that have been well-studied for compressing DNNs to learn compact representations, they suffer from non-negligible performance degradation in practice. In this paper, we propose Scalable Tensorizing Networks (STN), which dynamically and adaptively adjust the model size and decomposition structure without retraining. First, we account for compression during training by adding a low-rank regularizer to guarantee networks' desired low-rank characteristics in full tensor format. Then, considering network layers exhibit various low-rank structures, STN is obtained by a data-driven adaptive TD approach, for which the topological structure of decomposition per layer is learned from the pre-trained model, and the ranks are selected appropriately under specified storage constraints. As a result, STN is compatible with arbitrary network architectures and achieves higher compression performance and flexibility over other tensorizing versions. Comprehensive experiments on several popular architectures and benchmarks substantiate the superiority of our model towards improving parameter efficiency.

cs.CV↗

arXiv2022

Incoherent phonon transport dominates heat conduction across van der Waals superlattices

Heat conduction mechanisms in superlattices could be different across different types of interfaces. Van der Waals superlattices are structures physically assembled through weak van der Waals interactions by design, and may host properties beyond the traditional limits of lattice matching and processing compatibility, offering new types of interfaces. In this work, natural van der Waals (SnS)1.17(NbS2)n superlattices are synthesized, and their thermal conductivities are measured by time-domain thermoreflectance as a function of interface density. Our results show that heat conduction of (SnS)1.17(NbS2)n superlattices is dominated by interface scattering when the coherent length of phonons is larger than the superlattice period, indicating incoherent phonon transport dominates cross-plane heat conduction in van der Waals superlattices even when the period is atomically thin and abrupt. Moreover, our result suggests that the widely accepted heat conduction mechanism for conventional superlattices that coherent phonons dominate when the period is short, is not applicable due to symmetry breaking in most van der Waals superlattices. Our findings provide new insight for understanding the thermal behavior of van der Waals superlattices, and devise approaches for effective thermal management of superlattices depending on the distinct types of interfaces.

cond-mat.mtrl-sci↗

arXiv2021

Inverse obstacle scattering for elastic waves in the time domain

This paper concerns an inverse elastic scattering problem which is to determine a rigid obstacle from time domain scattered field data for a single incident plane wave. By using Helmholtz decomposition, we reduce the initial-boundary value problem of the time domain Navier equation to a coupled initial-boundary value problem of wave equations, and prove the uniqueness of the solution for the coupled problem by employing energy method. The retarded single layer potential is introduced to establish the coupled boundary integral equations, and the uniqueness is discussed for the solution of the coupled boundary integral equations. Based on the convolution quadrature method for time discretization, the coupled boundary integral equations are reformulated into a system of boundary integral equations in s-domain, and then a convolution quadrature based nonlinear integral equation method is proposed for the inverse problem. Numerical experiments are presented to show the feasibility and effectiveness of the proposed method.

math.NA↗

arXiv2015

The recoil correction and spin-orbit force for the possible $B^* \bar{B}^{}$ and $D^ \bar{D}^{*}$ states

In the framework of the one-boson exchange model, we have calculated the effective potentials between two heavy mesons $B^* \bar{B}^{*}$ and $D^* \bar{D}^{*}$ from the t- and u-channel $π$-, $η$-, $ρ$-, $ω$- and $σ$-meson exchanges. We keep the recoil corrections to the $B^* \bar{B}^{*}$ and $D^* \bar{D}^{*}$ systems up to $O(\frac{1}{M^2})$, which turns out to be important for the very loosely bound molecular states. Our numerical results show that the momentum-related corrections are favorable to the formation of the molecular states in the $I^G=1^+$, $J^{PC}=1^{+-}$ in the $B^* \bar{B}^{*}$ and $D^* \bar{D}^{*}$ systems.

hep-ph↗

arXiv2014

Hidden-Charm Tetraquarks and Charged Zc States

Experimentally several charged axial-vector hidden-charm states were reported. Within the framework of the color-magnetic interaction, we have systematically considered the mass spectrum of the hidden-charm and hidden-bottom tetraquark states. It is impossible to accommodate all the three charged states $Z_c(3900)$, $Z_c(4025)$ and $Z_c(4200)$ within the axial vector tetraquark spectrum simultaneously. Not all these three states are tetraquark candidates. Moreover, the eigenvector of the chromomagnetic interaction contains valuable information of the decay pattern of the tetraquark states. The dominant decay mode of the lowest axial vector tetraquark state is $J/ψπ$ while its $D^*\bar{D}$ and $\bar{D}^*D^*$ modes are strongly suppressed, which is in contrast with the fact that the dominant decay mode of $Z_c(3900)$ and $Z_c(4025)$ is $\bar{D}D^*$ and $\bar{D}^*D^*$ respectively. We emphasize that all the available experimental information indicates that $Z_c(4200)$ is a very promising candidate of the lowest axial vector hidden-charm tetraquark state.

hep-ph↗