arXiv subjects

Ziming Wang

Publications and source records attributed to Ziming Wang.

At least 55 records · Page 3Linked to original sources

arXiv2026

ICPRL: Acquiring Physical Intuition from Interactive Control

VLMs excel at static perception but falter in interactive reasoning in dynamic physical environments, which demands planning and adaptation to dynamic outcomes. Existing physical reasoning methods often depend on abstract symbolic inputs or lack the ability to learn and adapt from direct, pixel-based visual interaction in novel scenarios. We introduce ICPRL (In-Context Physical Reinforcement Learning), a framework inspired by In-Context Reinforcement Learning (ICRL) that empowers VLMs to acquire physical intuition and adapt their policies in-context. Our approach trains a vision-grounded policy model via multi-turn Group Relative Policy Optimization (GRPO) over diverse multi-episode interaction histories. This enables the agent to adapt strategies by conditioning on past trial-and-error sequences, without requiring any weight updates. This adaptive policy works in concert with a separately trained world model that provides explicit physical reasoning by predicting the results of potential actions. At inference, the policy proposes candidate actions, while the world model predicts outcomes to guide a root-node PUCT search to select the most promising action. Evaluated on the diverse physics-based puzzle-solving tasks in the DeepPHY benchmark, ICPRL demonstrates significant improvements across both its I. policy-only, and II. world-model-augmented stages. Notably, these gains are retained in unseen physical environments, demonstrating that our framework facilitates genuine in-context acquisition of the environment's physical dynamics from interactive experience.

cs.LG↗

arXiv2026

Procedural Fairness in Machine Learning

Fairness in machine learning (ML) has garnered significant attention. However, current research has mainly concentrated on the distributive fairness of ML models, with limited focus on another dimension of fairness, i.e., procedural fairness. In this paper, we first define the procedural fairness of ML models by drawing from the established understanding of procedural fairness in philosophy and psychology fields, and then give formal definitions of individual and group procedural fairness. Based on the proposed definition, we further propose a novel metric to evaluate the group procedural fairness of ML models, called $GPF_{FAE}$, which utilizes a widely used explainable artificial intelligence technique, namely feature attribution explanation (FAE), to capture the decision process of ML models. We validate the effectiveness of $GPF_{FAE}$ on a synthetic dataset and eight real-world datasets. Our experimental studies have revealed the relationship between procedural and distributive fairness of ML models. After validating the proposed metric for assessing the procedural fairness of ML models, we then propose a method for identifying the features that lead to the procedural unfairness of the model and propose two methods to improve procedural fairness based on the identified unfair features. Our experimental results demonstrate that we can accurately identify the features that lead to procedural unfairness in the ML model, and both of our proposed methods can significantly improve procedural fairness while also improving distributive fairness, with a slight sacrifice on the model performance.

cs.LG↗

arXiv2026

AllMem: A Memory-centric Recipe for Efficient Long-context Modeling

Large Language Models (LLMs) encounter significant performance bottlenecks in long-sequence tasks due to the computational complexity and memory overhead inherent in the self-attention mechanism. To address these challenges, we introduce \textsc{AllMem}, a novel and efficient hybrid architecture that integrates Sliding Window Attention (SWA) with non-linear Test-Time Training (TTT) memory networks. \textsc{AllMem} enables models to effectively scale to ultra-long contexts while mitigating catastrophic forgetting. This approach not only overcomes the representation constraints typical of linear memory models but also significantly reduces the computational and memory footprint during long-sequence inference. Furthermore, we implement a Memory-Efficient Fine-Tuning strategy to replace standard attention layers in pre-trained models with memory-augmented sliding window layers. This framework facilitates the efficient transformation of any off-the-shelf pre-trained LLM into an \textsc{AllMem}-based architecture. Empirical evaluations confirm that our 4k window model achieves near-lossless performance on 37k LongBench with a marginal 0.83 drop compared to full attention. Furthermore, on InfiniteBench at a 128k context, our 8k window variant outperforms full attention, which validates the effectiveness of our parameterized memory in mitigating noise and maintaining robust long-range modeling without the prohibitive costs of global attention.

cs.AI↗

arXiv2026

Equivariant Flow Matching for Point Cloud Assembly

The goal of point cloud assembly is to reconstruct a complete 3D shape by aligning multiple point cloud pieces. This work presents a novel equivariant solver for assembly tasks based on flow matching models. We first theoretically show that the key to learning equivariant distributions via flow matching is to learn related vector fields. Based on this result, we propose an assembly model, called equivariant diffusion assembly (Eda), which learns related vector fields conditioned on the input pieces. We further construct an equivariant path for Eda, which guarantees high data efficiency of the training process. Our numerical results show that Eda is highly competitive on practical datasets, and it can even handle the challenging situation where the input pieces are non-overlapped.

cs.CV↗

arXiv2026

Electrically Accessible Metamagnetic Transition via a Doping-Induced Low-Energy Magnetic State in Antiferromagnetic Insulator RFeO3

Low-energy antiferromagnetic phase transitions offer an appealing platform for low-power spintronic functionalities, yet their direct electrical access in insulating antiferromagnets remains challenging, particularly in the low-field regime where subtle Neeel vector reorientations dominate. Here, we demonstrate that targeted rare-earth-site engineering enables an electrically accessible metamagnetic transition in the insulating orthoferrite Ho0.5Dy0.5FeO3. By combining the distinct spin-reorientation sequences of DyFeO3 and HoFeO3, Dy substitution stabilizes a dual spin-reorientation pathway, hosting an intermediate state with a reduced energy barrier. This low-energy antiferromagnetic state can be tuned into the weak-ferromagnetic state under low magnetic fields. The critical field decreases with increasing temperature, providing a favorable window for functional manipulation. Both longitudinal and transverse spin Hall magnetoresistance channels exhibit clear and reproducible signatures of the metamagnetic transitions. Owing to the enhanced sensitivity of the transverse channel, additional low-field features are resolved, reflecting the projection of the Neel vector onto the spin-accumulation direction. Electrical transport measurements correlate directly with the magnetically determined phase boundaries, establishing a purely electrical access to low-energy phase transitions and to illustrate a viable pathway for exploring low-power spin dynamics in insulating oxide antiferromagnets.

cond-mat.mtrl-sci↗

arXiv2026

Global Compression Commander: Plug-and-Play Inference Acceleration for High-Resolution Large Vision-Language Models

Large vision-language models (LVLMs) excel at visual understanding, but face efficiency challenges due to quadratic complexity in processing long multi-modal contexts. While token compression can reduce computational costs, existing approaches are designed for single-view LVLMs and fail to consider the unique multi-view characteristics of high-resolution LVLMs with dynamic cropping. Existing methods treat all tokens uniformly, but our analysis reveals that global thumbnails can naturally guide the compression of local crops by providing holistic context for informativeness evaluation. In this paper, we first analyze dynamic cropping strategy, revealing both the complementary nature between thumbnails and crops, and the distinctive characteristics across different crops. Based on our observations, we propose ``Global Compression Commander'' (\textit{i.e.}, \textbf{GlobalCom$^2$}), a novel plug-and-play token compression framework for HR-LVLMs. GlobalCom$^2$ leverages thumbnail as the ``commander'' to guide the compression of local crops, adaptively preserving informative details while eliminating redundancy. Extensive experiments show that GlobalCom$^2$ maintains over \textbf{90\%} performance while compressing \textbf{90\%} visual tokens, reducing FLOPs and peak memory to \textbf{9.1\%} and \textbf{60\%}.

cs.CV↗

arXiv2026

CombatVLA: An Efficient Vision-Language-Action Model for Combat Tasks in 3D Action Role-Playing Games

Recent advances in Vision-Language-Action models (VLAs) have expanded the capabilities of embodied intelligence. However, significant challenges remain in real-time decision-making in complex 3D environments, which demand second-level responses, high-resolution perception, and tactical reasoning under dynamic conditions. To advance the field, we introduce CombatVLA, an efficient VLA model optimized for combat tasks in 3D action role-playing games(ARPGs). Specifically, our CombatVLA is a 3B model trained on video-action pairs collected by an action tracker, where the data is formatted as action-of-thought (AoT) sequences. Thereafter, CombatVLA seamlessly integrates into an action execution framework, allowing efficient inference through our truncated AoT strategy. Experimental results demonstrate that CombatVLA not only outperforms all existing models on the combat understanding benchmark but also achieves a 50-fold acceleration in game combat. Moreover, it has a higher task success rate than human players. We will open-source all resources, including the action tracker, dataset, benchmark, model weights, training code, and the implementation of the framework at https://combatvla.github.io/.

cs.CV↗

arXiv2026

Anatomy of parameter-estimation biases in overlapping gravitational-wave signals: detector network

With the significantly improved sensitivity and a wider frequency band, the next-generation gravitational-wave (GW) detectors are anticipated to detect $\sim 10^5$ GW signals per year with durations from hours to days, leading to inevitable signal overlaps in the data stream. While a direct fitting for all signals may be challenging, extracting only one signal will be biased by its overlap with other signals. From this perspective, understanding how the biases arise from the overlapping and their dependence on the signal parameters is crucial for developing effective algorithms. In this work, we extend the anatomy of biases in single-detector cases (Wang et al. 2024) to a detector network. Specifically, we examine how the biases of the chirp mass, symmetric mass ratio, luminosity distance, and coalescence time depend on the source's sky position and orientation, as well as on the coalescence time and phase. We propose a new quantity, named the bias integral, as a useful tool, and establish relationship between the biases in a single detector and that in the entire network, with explicit dependence on extrinsic parameters. Using a 3-detector network as an example, we further explore the potential of a network to suppress biases due to the detectors' different locations and orientations. We find that location generally has a smaller effect than orientation, and becomes significant only when the time separation between signals is below sub-seconds. Through a population-level simulation over the extrinsic parameters, we find that nearly half of overlapping signals will lead to larger biases in the network compared to a single detector, highlighting the need to cope with overlapping biases in a detector network.

gr-qc↗

arXiv2026

Neural-network-based Self-triggered Observed Platoon Control for Autonomous Vehicles

This paper investigates autonomous vehicle (AV) platoon control under uncertain dynamics and intermittent communication, which remains a critical challenge in intelligent transportation systems. To address these issues, this paper proposes an adaptive consensus tracking control framework for nonlinear multi-agent systems (MASs). The proposed approach integrates backstepping design, a nonlinear sampled-data observer, radial basis function neural networks, and a self-triggered communication mechanism. The radial basis function neural networks approximate unknown nonlinearities and time-varying disturbances, thereby enhancing system robustness. A distributed observer estimates neighboring states based on limited and intermittent measurements, thereby reducing dependence on continuous communication. Moreover, self-triggered mechanism is developed to determine triggering instants, guaranteeing a strictly positive minimum inter-event time and preventing Zeno behavior. The theoretical analysis proves that all closed-loop signals are uniformly ultimately bounded (UUB), and tracking errors converge to a compact set. Simulation results demonstrate that the proposed approach achieves high robustness, adaptability, and communication efficiency, making it suitable for real-world networked vehicle systems.

eess.SY↗

arXiv2025

Lightweight posterior construction for gravitational-wave catalogs with the Kolmogorov-Arnold network

Neural density estimation has seen widespread applications in the gravitational-wave (GW) data analysis, which enables real-time parameter estimation for compact binary coalescences and enhances rapid inference for subsequent analysis such as population inference. In this work, we explore the application of using the Kolmogorov-Arnold network (KAN) to construct efficient and interpretable neural density estimators for lightweight posterior construction of GW catalogs. By replacing conventional activation functions with learnable splines, KAN achieves superior interpretability, higher accuracy, and greater parameter efficiency on related scientific tasks. Leveraging this feature, we propose a KAN-based neural density estimator, which ingests megabyte-scale GW posterior samples and compresses them into model weights of tens of kilobytes. Subsequently, analytic expressions requiring only several kilobytes can be further distilled from these neural network weights with minimal accuracy trade-off. In practice, GW posterior samples with fidelity can be regenerated rapidly using the model weights or analytic expressions for subsequent analysis. Our lightweight posterior construction strategy is expected to facilitate user-level data storage and transmission, paving a path for efficient analysis of numerous GW events in the next-generation GW detectors.

gr-qc↗

arXiv2025

Fine-Tuned LLMs Know They Don't Know: A Parameter-Efficient Approach to Recovering Honesty

The honesty of Large Language Models (LLMs) is increasingly important for safe deployment in high-stakes domains. However, this crucial trait is severely undermined by supervised fine-tuning (SFT), a common technique for model specialization. Existing recovery methods rely on data-intensive global parameter adjustments, implicitly assuming that SFT deeply corrupts the models' ability to recognize their knowledge boundaries. However, we observe that fine-tuned LLMs still preserve this ability; what is damaged is their capacity to faithfully express that awareness. Building on this, we propose Honesty-Critical Neurons Restoration (HCNR) to surgically repair this suppressed capacity. HCNR identifies and restores key expression-governing neurons to their pre-trained state while harmonizing them with task-oriented neurons via Hessian-guided compensation. Experiments on four QA tasks and five LLM families demonstrate that HCNR effectively recovers 33.25% of the compromised honesty while achieving at least 2.23x speedup with over 10x less data compared to baseline methods, offering a practical solution for trustworthy LLM deployment.

cs.CL↗

arXiv2025

Who Grants the Agent Power? Defending Against Instruction Injection via Task-Centric Access Control

AI agents capable of GUI understanding and Model Context Protocol are increasingly deployed to automate mobile tasks. However, their reliance on over-privileged, static permissions creates a critical vulnerability: instruction injection. Malicious instructions, embedded in otherwise benign content like emails, can hijack the agent to perform unauthorized actions. We present AgentSentry, a lightweight runtime task-centric access control framework that enforces dynamic, task-scoped permissions. Instead of granting broad, persistent permissions, AgentSentry dynamically generates and enforces minimal, temporary policies aligned with the user's specific task (e.g., register for an app), revoking them upon completion. We demonstrate that AgentSentry successfully prevents an instruction injection attack, where an agent is tricked into forwarding private emails, while allowing the legitimate task to complete. Our approach highlights the urgent need for intent-aligned security models to safely govern the next generation of autonomous agents.

cs.CR↗

arXiv2025

Who Moved My Transaction? Uncovering Post-Transaction Auditability Vulnerabilities in Modern Super Apps

Super apps are the cornerstones of modern digital life, embedding financial transactions into nearly every aspect of daily routine. The prevailing security paradigm for these platforms is overwhelmingly focused on pre-transaction authentication, preventing unauthorized payments before they occur. We argue that a critical vulnerability vector has been largely overlooked: the fragility of post-transaction audit trails. We investigate the ease with which a user can permanently erase their transaction history from an app's interface, thereby concealing unauthorized or sensitive activities from the account owner. To quantify this threat, we conducted an empirical study with 6 volunteers who performed a cross-evaluation on six super apps. Our findings are alarming: all six applications studied allow users to delete transaction records, yet a staggering five out of six (83+\%) fail to protect these records with strong authentication. Only one app in our study required biometric verification for deletion. This study provides the first concrete evidence of this near-ubiquitous vulnerability, demonstrating a critical gap in the current mobile security landscape and underscoring the urgent need for a paradigm shift towards ensuring post-transaction audit integrity.

cs.CR↗

arXiv2025

A Common Synchrotron Origin for Prompt Gamma-Ray and Soft X-Ray Emission in GRBs: Evidence from Joint Spectral Analysis

The recent launches of the Einstein Probe (EP) and the Space Variable Objects Monitor (SVOM) mission have led to the detection of a growing number of long GRBs with significant, early soft X-ray flux during their gamma-ray emission, prompting the question of whether their multi-band prompt emission shares a common origin in region and mechanism. To address this, we utilize the 20-year Swift archival data, which provides a substantial sample of joint soft X-ray and gamma-ray observations, enabling a systematic joint spectral study. We resolve 110 temporal pulses from 46 GRBs and find that a single power-law model with a low-energy break or cutoff adequately describes the prompt spectra from 150 keV down to 0.5 keV. More than half of the sample pulses require a break around a few keV, with average spectral indices $\langle α_1 \rangle = -0.88$ and $\langle α_2 \rangle = -1.46$ consistent with synchrotron radiation in a marginally fast-cooling regime. The observed spectral evolution and the distribution of indices support a single-emission-region origin, where the varying spectral shapes are largely governed by the evolution of the synchrotron cooling frequency $ν_c$ and the effect of finite emission width. The observed differences in the temporal behavior between X-ray and gamma-ray light curves can be naturally explained by this spectral evolution across the broad band.

astro-ph.HE↗

arXiv2025

Ringdown mode amplitudes of charged binary black holes

The ringdown phase of the binary black hole (BBH) merger provides a clean and direct probe of strong-field gravity and tests of the nature of black holes. The quasinormal mode (QNM) frequencies in modified gravity theories, as well as their amplitudes and phases, might deviate from the Kerr ones in general relativity. Charged black holes (BHs) in Einstein-Maxwell theory provide an excellent example of a beyond-Kerr solution with direct astrophysical and fundamental physics applications. In this work, we extract the ringdown mode amplitudes and phases for charged BBH mergers based on fully general relativistic simulations with charge-to-mass ratio up to 0.3. Our results suggest that even though the inspiral phase of charged BBHs can be significantly accelerated or decelerated, the ringdown mode excitation only changes mildly. We further explore the charge detectability with the ringdown-only signal for the Einstein Telescope and Cosmic Explorer. We find that previous studies may have overestimated the charge detectability and including higher modes in charged waveforms is necessary for future ringdown analysis. This constitutes the first such analysis based on waveforms generated by numerical relativity simulations of charged BHs in full Einstein-Maxwell theory.

gr-qc↗

arXiv2025

Fixed-Time Voltage Regulation for Boost Converters via Unit-Safe Saturating Functions

This paper explores the voltage regulation challenges in boost converter systems, which are critical components in power electronics due to their ability to step up voltage levels efficiently. The proposed control algorithm ensures fixed-time stability, a desirable property that guarantees system stability within a fixed time frame regardless of initial conditions. To tackle the common chattering issues in conventional fixed-time control methods, a novel class of function families is introduced. State observers and adaptive parameters are utilized to manage the uncertainties associated with unknown load resistance. Furthermore, a new disturbance observer is developed using the proposed function family, and its advantages and limitations are illustrated through comparison with existing designs. Finally, both non-real-time and real-time simulations are conducted to validate the effectiveness and deployability of the proposed control algorithm.

eess.SY↗

arXiv2025

Rigorous analytic solution to the gravitational-wave overlapping event rates

In the era of the next-generation gravitational-wave detectors, signal overlaps will become prevalent due to high detection rate and long signal duration, posing significant challenges to data analysis. While effective algorithms are being developed, there still lacks an integrated understanding on the statistical properties for the population of overlapping compact-binary-coalescence signals. For the first time, in order to aid rapid and robust estimation, we rigorously derive and establish analytical expressions for the expectation and variance for the number of overlapping events. This framework is highly extensible, allowing analytical calculation for more complicated scenarios, such as multi-signal overlaps, overlaps between different types of sources, and source-dependent thresholds. We also mathematically prove that the time difference between events in a single observation run is described by the beta distribution, offering an analytical prior reference for Bayesian analysis.

gr-qc↗

arXiv2025

DeepPHY: Benchmarking Agentic VLMs on Physical Reasoning

Although Vision Language Models (VLMs) exhibit strong perceptual abilities and impressive visual reasoning, they struggle with attention to detail and precise action planning in complex, dynamic environments, leading to subpar performance. Real-world tasks typically require complex interactions, advanced spatial reasoning, long-term planning, and continuous strategy refinement, usually necessitating understanding the physics rules of the target scenario. However, evaluating these capabilities in real-world scenarios is often prohibitively expensive. To bridge this gap, we introduce DeepPHY, a novel benchmark framework designed to systematically evaluate VLMs' understanding and reasoning about fundamental physical principles through a series of challenging simulated environments. DeepPHY integrates multiple physical reasoning environments of varying difficulty levels and incorporates fine-grained evaluation metrics. Our evaluation finds that even state-of-the-art VLMs struggle to translate descriptive physical knowledge into precise, predictive control.

cs.AI↗