arXiv subjects

Zhe Ji

Publications and source records attributed to Zhe Ji.

16 recordsLinked to original sources

arXiv2026

Embodied-BenchForge: A Closed-Loop Agentic Workflow for Embodied Benchmark Construction

Agentic systems offer a promising way to automate embodied benchmark construction, but existing approaches typically cover isolated stages or remain specialized to predefined environments and task families. More importantly, multi-step construction produces dependent intermediate artifacts that are often passed downstream without artifact-specific verification, allowing local defects to propagate into the final benchmark. We present Embodied-BenchForge, an agentic framework that transforms user-specified evaluation intents into complete embodied benchmark artifacts. It formulates construction as Closed-Loop Benchmark Synthesis, integrating forward artifact synthesis with backward verification and repair. Skill-Orchestrated Artifact Synthesis composes typed and reusable skills into executable workflows, while an artifact dependency graph records intermediate outputs and their dependencies. Requirement-Guided Verification and Repair applies artifact-specific contracts throughout construction and uses provenance to trigger local re-execution or upstream rollback when verification fails. Embodied-BenchForge constructs six benchmarks covering diverse embodied scenarios in the Offline EQA Track, together with one interactive benchmark containing 220 executable tasks in the Interactive Embodied Track. Evaluations of representative MLLMs and embodied agents show that the benchmarks distinguish model capabilities in both observation-based understanding and closed-loop execution. Quality assessment and ablations validate benchmark quality and the effectiveness of verification and repair, while repair and skill-reuse analyses demonstrate efficient localized recovery and cross-benchmark reusability.

cs.AI↗

arXiv2026

Embodied-BenchClaw: An Autonomous Multi-Agent System for Embodied Spatial Intelligence Benchmark Construction

Benchmarks are essential for evaluating embodied spatial intelligence, yet their construction is labor-intensive, hard to reuse, and difficult to maintain. Existing embodied benchmarks are often static and may quickly become saturated as models improve, limiting their ability to distinguish new capabilities. We propose Embodied-BenchClaw, an autonomous agentic system for constructing embodied spatial intelligence benchmarks. Given a user-specified evaluation intent, Embodied-BenchClaw automatically produces a complete and continually updatable benchmark package through a five-stage pipeline: intent blueprinting, data collection, structuring and cleaning, benchmark synthesis, and evaluation reporting. The pipeline is coordinated by three agents for planning, construction, and evaluation. To improve reusability and reliability, Embodied-BenchClaw introduces an extensible Skill Library and process quality control, enabling benchmark construction to be composable, verifiable, and repairable. We instantiate multiple benchmarks covering indoor spatial reasoning, outdoor spatial reasoning, robotic manipulation, quadruped robot navigation, UAV/aerial-view understanding, and static benchmark enhancement. These benchmarks span diverse embodied carriers, data sources, and spatial capabilities. Experiments with human evaluation, judge-based assessment, consistency checks, cost analysis, and ablations show that Embodied-BenchClaw can construct verifiable, executable, maintainable, and diagnostically useful embodied spatial benchmarks with reduced manual effort.

cs.AI↗

arXiv2026

Multi-needle Localization for Pelvic Seed Implant Brachytherapy based on Tip-handle Detection and Matching

Accurate multi-needle localization in intraoperative CT images is crucial for optimizing seed placement in pelvic seed implant brachytherapy. However, this task is challenging due to poor image contrast and needle adhesion. This paper presents a novel approach that reframes needle localization as a tip-handle detection and matching problem to overcome these difficulties. An anchor-free network, based on HRNet, is proposed to extract multi-scale features and accurately detect needle tips and handles by predicting their centers and orientations using decoupled branches for heatmap regression and polar angle prediction. To associate detected tips and handles into individual needles, a greedy matching and merging (GMM) method designed to solve the unbalanced assignment problem with constraints (UAP-C) is presented. The GMM method iteratively selects the most probable tip-handle pairs and merges them based on a distance metric to reconstruct 3D needle paths. Evaluated on a dataset of 100 patients, the proposed method demonstrates superior performance, achieving higher precision and F1 score compared to a segmentation-based method utilizing the nnUNet model,thereby offering a more robust and accurate solution for needle localization in complex clinical scenarios.

cs.CV↗

arXiv2026

Seamless Handover in Direct-to-Device Satellite Networks: From an Interference-Aware Perspective

The direct-to-device (D2D) satellite network is an important 6G evolution direction to enable seamless ubiquitous connectivity. However, the network faces critical handover challenges due to high satellite mobility and wide beam footprints. Conventional handover strategies, mostly designed for terrestrial networks, may encounter excessive co-channel interference (CCI) and service degradation in the satellite environments. To address the issue, this paper introduces a novel handover optimization method to perform dynamic adjustment of an important parameter called elevation angle threshold (EAT) from an interference-aware perspective. Explicitly, we first analyze the trade-off between satellite visibility and CCI. Then, we propose a numerical algorithm to determine the optimal EAT that can achieve seamless coverage with CCI. We validate our method using a customized D2D LEO satellite network simulator in the Network Simulator (NS-3). The results demonstrate that the EAT optimization significantly reduces packet loss and hence enhances handover reliability. The work highlights the importance of interference-aware handover design for improving service continuity in future D2D satellite networks.

cs.NI↗

arXiv2025

An Iterative LLM Framework for SIBT utilizing RAG-based Adaptive Weight Optimization

Seed implant brachytherapy (SIBT) is an effective cancer treatment modality; however, clinical planning often relies on manual adjustment of objective function weights, leading to inefficiencies and suboptimal results. This study proposes an adaptive weight optimization framework for SIBT planning, driven by large language models (LLMs). A locally deployed DeepSeek-R1 LLM is integrated with an automatic planning algorithm in an iterative loop. Starting with fixed weights, the LLM evaluates plan quality and recommends new weights in the next iteration. This process continues until convergence criteria are met, after which the LLM conducts a comprehensive evaluation to identify the optimal plan. A clinical knowledge base, constructed and queried via retrieval-augmented generation (RAG), enhances the model's domain-specific reasoning. The proposed method was validated on 23 patient cases, showing that the LLM-assisted approach produces plans that are comparable to or exceeding clinically approved and fixed-weight plans, in terms of dose homogeneity for the clinical target volume (CTV) and sparing of organs at risk (OARs). The study demonstrates the potential use of LLMs in SIBT planning automation.

physics.med-ph↗

arXiv2025

Affine Invariant Semi-Blind Receiver: Joint Channel Estimation and High-Order Signal Detection for Multiuser Massive MIMO-OFDM Systems

Massive multiple input and multiple output (MIMO) systems with orthogonal frequency division multiplexing (OFDM) are foundational for downlink multi-user (MU) communication in future wireless networks, for their ability to enhance spectral efficiency and support a large number of users simultaneously. However, high user density intensifies severe inter-user interference (IUI) and pilot overhead. Consequently, existing blind and semi-blind channel estimation (CE) and signal detection (SD) algorithms suffer performance degradation and increased complexity, especially when further challenged by frequency-selective channels and high-order modulation demands. To this end, this paper proposes a novel semi-blind joint channel estimation and signal detection (JCESD) method. Specifically, the proposed approach employs a hybrid precoding architecture to suppress IUI. Furthermore we formulate JCESD as a non-convex constellation fitting optimization exploiting constellation affine invariance. Few pilots are used to achieve coarse estimation for initialization and ambiguity resolution. For high-order modulations, a data augmentation mechanism utilizes the symmetry of quadrature amplitude modulation (QAM) constellations to increase the effective number of samples. To address frequency-selective channels, CE accuracy is then enhanced via an iterative refinement strategy that leverages improved SD results. Simulation results demonstrate an average throughput gain of 11\% over widely used pilot-based methods in MU scenarios, highlighting the proposed method's potential to improve spectral efficiency.

eess.SP↗

arXiv2025

A large-scale distributed parallel discrete event simulation engines based on Warped2 for Wargaming simulation

Rising demand for complex simulations highlights conventional engines'scalability limits, spurring Parallel Discrete Event Simulation (PDES) adoption.Warped2, a PDES engine leveraging Time Warp synchronization with Pending Event Set optimization, delivers strong performance, it struggles with inherent wargaming limitations: inefficient LP resource allocation during synchronization and unaddressed complex entity interaction patterns. To address these challenges, we present an optimized framework featuring four synergistic improvements: (1) Asynchronous listener threads are introduced to address event monitoring latency in large-scale scenarios, instead of synchronous polling mechanisms, (2) METIS-based load rebalancing strategy is incorporated to address the issue of dynamic event allocation during real-world simulation, (3) Entity interaction solver with constraint satisfaction mechanisms is designed to mitigate state conflicts, and (4) Spatial hashing algorithm to overcome O(n^2) complexity bottlenecks in large-scale nearest-neighbor searches. Experimental validation through a GridWorld demo demonstrates significant enhancements in temporal fidelity and computational efficiency. Benchmark results show our framework achieves 16x acceleration over baseline implementations and maintains 8x speedup over 1-thread configuration across MPI and Pthreads implementations.The combined load balancing and LP migration strategy reduces synchronization overhead by 58.18%, with load balancing accounting for 57% of the total improvement as the dominant optimization factor. These improvements provide an enhanced solution for PDES implementation in large-scale simulation scenarios.

cs.DC↗

arXiv2025

A feature-preserving parallel particle generation method for complex geometries

In this paper, a Feature-preserving Particle Generation (FPPG) method for arbitrary complex geometry is proposed. Instead of basing on implicit geometries, such as level-set, FPPG employs an explicit geometric representation for the parallel and automatic generation of high-quality surface and volume particles, which enables the full preservation of geometric features, such as sharp edges, singularities and etc. Several new algorithms are proposed in this paper to achieve the aforementioned objectives. First, a particle mapping and feature line extraction algorithm is proposed to ensure the adequate representation of arbitrary complex geometry. An improved and efficient data structure is developed too to maximize the parallel efficiency and to optimize the memory footprint. Second, the physics-based particle relaxation procedure is tailored for the explicit geometric representation to achieve a uniform particle distribution. Third, in order to handle large-scale industrial models, the proposed FPPG method is entirely parallelized on shared memory systems and Boolean operations are allowed to tackle structures with multiple assemblies. Intensive numerical tests are carried out to demonstrate the capabilities of FPPG. The scalability tests show that a speedup of ~10X is achieved through multi-threading parallelization with various models. Comparative studies with other particle generation methods show that FPPG achieves better performance in both runtime and accuracy. Last, two industrial cases of vehicle wading and gearbox oiling are studied to illustrate that FPPG is applicable to complex geometries.

physics.comp-ph↗

arXiv2024

Curator Attack: When Blackbox Differential Privacy Auditing Loses Its Power

A surge in data-driven applications enhances everyday life but also raises serious concerns about private information leakage. Hence many privacy auditing tools are emerging for checking if the data sanitization performed meets the privacy standard of the data owner. Blackbox auditing for differential privacy is particularly gaining popularity for its effectiveness and applicability to a wide range of scenarios. Yet, we identified that blackbox auditing is essentially flawed with its setting: small probabilities or densities are ignored due to inaccurate observation. Our argument is based on a solid false positive analysis from a hypothesis testing perspective, which is missed out by prior blackbox auditing tools. This oversight greatly reduces the reliability of these tools, as it allows malicious or incapable data curators to pass the auditing with an overstated privacy guarantee, posing significant risks to data owners. We demonstrate the practical existence of such threats in classical differential privacy mechanisms against four representative blackbox auditors with experimental validations. Our findings aim to reveal the limitations of blackbox auditing tools, empower the data owner with the awareness of risks in using these tools, and encourage the development of more reliable differential privacy auditing methods.

cs.CR↗

arXiv2024

Weights Shuffling for Improving DPSGD in Transformer-based Models

Differential Privacy (DP) mechanisms, especially in high-dimensional settings, often face the challenge of maintaining privacy without compromising the data utility. This work introduces an innovative shuffling mechanism in Differentially-Private Stochastic Gradient Descent (DPSGD) to enhance the utility of large models at the same privacy guarantee of the unshuffled case. Specifically, we reveal that random shuffling brings additional randomness to the trajectory of gradient descent while not impacting the model accuracy by the permutation invariance property -- the model can be equivalently computed in both forward and backward propagations under permutation. We show that permutation indeed improves the privacy guarantee of DPSGD in theory, but tracking the exact privacy loss on shuffled model is particularly challenging. Hence we exploit the approximation on sum of lognormal distributions to derive the condition for the shuffled DPSGD to meet the DP guarantee. Auditing results show that our condition offers a DP guarantee quite close to the audited privacy level, demonstrating our approach an effective estimation in practice. Experimental results have verified our theoretical derivation and illustrate that our mechanism improves the accuracy of DPSGD over the state-of-the-art baselines on a variety of models and tasks.

cs.LG↗

arXiv2024

Crafter: Facial Feature Crafting against Inversion-based Identity Theft on Deep Models

With the increased capabilities at the edge (e.g., mobile device) and more stringent privacy requirement, it becomes a recent trend for deep learning-enabled applications to pre-process sensitive raw data at the edge and transmit the features to the backend cloud for further processing. A typical application is to run machine learning (ML) services on facial images collected from different individuals. To prevent identity theft, conventional methods commonly rely on an adversarial game-based approach to shed the identity information from the feature. However, such methods can not defend against adaptive attacks, in which an attacker takes a countermove against a known defence strategy. We propose Crafter, a feature crafting mechanism deployed at the edge, to protect the identity information from adaptive model inversion attacks while ensuring the ML tasks are properly carried out in the cloud. The key defence strategy is to mislead the attacker to a non-private prior from which the attacker gains little about the private identity. In this case, the crafted features act like poison training samples for attackers with adaptive model updates. Experimental results indicate that Crafter successfully defends both basic and possible adaptive attacks, which can not be achieved by state-of-the-art adversarial game-based methods.

cs.CR↗

arXiv2022

A class of new high-order finite-volume TENO schemes for hyperbolic conservation laws with unstructured meshes

The recently proposed high-order TENO scheme [Fu et al., Journal of Computational Physics, 305, pp.333-359] has shown great potential in predicting complex fluids owing to the novel weighting strategy, which ensures the high-order accuracy, the low numerical dissipation, and the sharp shock-capturing capability. However, the applications are still restricted to simple geometries with Cartesian or curvilinear meshes. In this work, a new class of high-order shock-capturing TENO schemes for unstructured meshes are proposed. Similar to the standard TENO schemes and some variants of WENO schemes, the candidate stencils include one large stencil and several small third-order stencils. Following a strong scale-separation procedure, a tailored novel ENO-like stencil selection strategy is proposed such that the high-order accuracy is restored in smooth regions by selecting the candidate reconstruction on the large stencil while the ENO property is enforced near discontinuities by adopting the candidate reconstruction from smooth small stencils. The nonsmooth stencils containing genuine discontinuities are explicitly excluded from the final reconstruction, leading to excellent numerical stability. Different from the WENO concept, such unique sharp stencil selection retains the low numerical dissipation without sacrificing the shock-capturing capability. The newly proposed framework enables arbitrarily high-order TENO reconstructions on unstructured meshes. For conceptual verification, the TENO schemes with third- to sixth-order accuracy are constructed. Without parameter tuning case by case, the performance of the proposed TENO schemes is demonstrated by examining a set of benchmark cases with broadband flow length scales.

math.NA↗

arXiv2020

A Feature-aware SPH for Isotropic Unstructured Mesh Generation

In this paper, we present a feature-aware SPH method for the concurrent and automated isotropic unstructured mesh generation. Two additional objectives are achieved with the proposed method compared to the original SPH-based mesh generator (Fu et al., 2019). First, a feature boundary correction term is introduced to address the issue of incomplete kernel support at the boundary vicinity. The mesh generation of feature curves, feature surfaces and volumes can be handled concurrently without explicitly following a dimensional sequence. Second, a two-phase model is proposed to characterize the mesh-generation procedure by a feature-size-adaptation phase and a mesh-quality-optimization phase. By proposing a new error measurement criterion and an adaptive control system with two sets of simulation parameters, the objectives of faster feature-size adaptation and local mesh-quality improvement are merged into a consistent framework. The proposed method is validated with a set of 2D and 3D numerical tests with different complexities and scales. The results demonstrate that high-quality meshes are generated with a significant speedup of convergence.

cs.GR↗

arXiv2019

A Consistent Parallel Isotropic Unstructured Mesh Generation Method based on Multi-phase SPH

In this paper, we propose a consistent parallel unstructured mesh generator based on a multi-phase SPH method. A set of physics-motivated modeling equations are developed to achieve the targets of domain decomposition, communication volume optimization and high-quality unstructured mesh generation simultaneously. A unified density field is defined as the target function for both partitioning the geometry and distributing the mesh-vertexes. A multi-phase Smoothing Particle Hydrodynamics (SPH) method is employed to solve the governing equations. All the optimization targets are achieved implicitly and consistently by the particle relaxation procedure without constructing triangulation/tetrahedralization explicitly. The target of communication reduction is achieved by introducing a surface tension model between distinct partitioning sub-domains, which are characterized by colored SPH particles. The resulting partitioning diagram features physically localized sub-domains and optimized interface communication. The target of optimizing the mesh quality is achieved by introducing a tailored equation-of-state (EOS) and a smooth isotropic kernel function. The mesh quality near the interface of neighboring sub-domains is improved by gradually removing the surface-tension force once a steady state is achieved. The proposed method is developed basing on a new parallel environment for multi-resolution SPH to exploit both coarse- and fine-grained parallelism. A set of benchmarks are conducted to verify that all the optimization targets are achieved consistently within the current framework.

physics.comp-ph↗

arXiv2018

A Lagrangian Inertial Centroidal Voronoi Particle method for dynamic load balancing in particle-based simulations

In this paper we develop a Lagrangian Inertial Centroidal Voronoi Particle (LICVP) method to extend the original CVP method \cite{fu2017physics} to dynamic load balancing in particle-based simulations. Two new concepts are proposed to address the additional problems encountered in repartitioning the system. First, a background velocity is introduced to transport Voronoi particle according to the local fluid field, which facilitates data reuse and lower data redistribution cost during rebalancing. Second, in order to handle problems with skew-aligned computational load and large void space, we develop an inertial-based partitioning strategy, where the inertial matrix is utilized to characterize the load distribution, and to confine the motion of Voronoi particles dynamically adapting to the physical simulation. Intensive numerical tests in fluid dynamics simulations reveal that the underlying LICVP method improves the incremental property remarkably without sacrifices on other objectives, i.e. the inter-processor communication is optimized simultaneously, and the repartitioning procedure is highly efficient.

physics.comp-ph↗

arXiv2016

Unconstrained Two-parallel-plane Model for Focused Plenoptic Cameras Calibration

The plenoptic camera can capture both angular and spatial information of the rays, enabling 3D reconstruction by single exposure. The geometry of the recovered scene structure is affected by the calibration of the plenoptic camera significantly. In this paper, we propose a novel unconstrained two-parallel-plane (TPP) model with 7 parameters to describe a 4D light field. By reconstructing scene points from ray-ray association, a 3D projective transformation is deduced to establish the relationship between the scene structure and the TPP parameters. Based on the transformation, we simplify the focused plenoptic camera as a TPP model and calibrate its intrinsic parameters. Our calibration method includes a close-form solution and a nonlinear optimization by minimizing re-projection error. Experiments on both simulated data and real scene data verify the performance of the calibration on the focused plenoptic camera.

cs.CV↗