arXiv subjects

Zheng Zhu

Publications and source records attributed to Zheng Zhu.

At least 19 recordsLinked to original sources

arXiv2026

HALO-WA: Hybrid-Attention Latent-Guided Online Reinforcement Learning for World-Action Models

World-action (WA) models can generate long-horizon action chunks for general-purpose robotic manipulation, but they remain vulnerable to calibration, perception, and contact-dynamics errors in real-world precision tasks, often failing in the final few millimeters of alignment or insertion. We propose HALO-WA, a hybrid-attention latent-guided online reinforcement learning (RL) framework for WA models, which leverages latent features and action priors from the WA generation process through a lightweight actor-critic adapter to enable fast online adaptation to real deployment errors. HALO-WA introduces a hybrid-attention structure that preserves the temporal consistency of action chunks while reading task-relevant information from WA latents conditioned on visual context and end-stage correction requirements, thereby producing refined action chunks. We validate HALO-WA on four real-world precision manipulation tasks, where it improves the average success rate from 26.4\% for WA-base to 87.1\%, outperforming the strongest baseline by 19.2 percentage points while requiring only 45--75 minutes of online training per task. To facilitate reproducibility, we further conduct supplementary simulation experiments in RoboTwin and release the code at https://github.com/YeanRoot/HALO-WA.

cs.RO↗

arXiv2026

GigaBrain-WBC-0.5: A Behavior World Model for Robust Humanoid Whole-Body Tracking with Environment Interaction

General-purpose motion trackers enable humanoid robots to follow diverse whole-body motions while maintaining balance, but are trained only on flat ground, failing to exploit bipedal mobility over complex terrain. Cross-terrain controllers, meanwhile, are task-specific or accept only low-dimensional locomotion commands. We introduce InterTrack, the first behavior world model (BWM) for robust whole-body tracking with environment interaction. Its Transformer jointly predicts the next action, state, and behavior distribution, learning environment-conditioned dynamics. To scale interaction training data, an automatic annotation pipeline reconstructs 3D support geometry from retargeted motions. At deployment, the policy handles commands implausible in the current environment in a "best-effort" manner. Quantitatively, InterTrack achieves an 81.3% success rate on terrain interaction (4.3 times the best evaluated baseline) and a 99.3% fall-recovery rate, while also improving free-space tracking and outperforming three leading tracking baselines across all of these regimes. To our knowledge, we provide the first demonstration of real-time cross-terrain whole-body teleoperation on a humanoid robot, alongside object interaction, stable responses to missing supports, and robust recovery from falls.

cs.RO↗

arXiv2026

Drive-HWM: Hierarchical World Models for Dynamic-Latent Guided Autonomous Driving

World models offer a promising paradigm for autonomous driving by predicting how traffic scenes may evolve and using such predictions to support action generation. However, existing approaches either separate future prediction from action generation or jointly predict them at the same temporal scale, making it difficult to simultaneously achieve long-horizon anticipation and responsive, observation-grounded decision making. We present Drive-HWM, a hierarchical slow--fast world modeling framework that organizes future representation prediction and action generation at complementary temporal scales. The slow world model predicts multi-step future representations to capture extended scene evolution. To explicitly model the abundant motion dynamics in driving environments, we introduce Dynamic-Aware Latents learned through optical-flow prediction. Guided by these future representations, the fast model uses a lightweight multimodal backbone and an autoregressive expert to jointly predict the next frame and the immediate action from the latest observation. Next-frame prediction encourages the fast model to capture imminent scene evolution, while one-step action generation allows decisions to be continuously updated as new observations arrive. Extensive experiments on NAVSIM v1 and v2 demonstrate the strong driving performance of Drive-HWM. Comprehensive ablation studies further validate the effectiveness of the hierarchical slow--fast design, dynamics-aware future representations, and joint next-frame and action prediction.

cs.CV↗

arXiv2026

XPolicyLab: A Unified Standard and Open Ecosystem for Robot Policy Evaluation and Deployment

Robot policy evaluation and deployment remain fragmented by model-specific software dependencies, data representations, and runtime interfaces, so that connecting N policies to M evaluation environments requires O(NM) separate integrations. We present XPolicyLab, a unified standard and open ecosystem that reduces this cost to O(N+M). XPolicyLab specifies common observation, action, and trajectory schemas together with a minimal adapter interface for observation updates, action prediction, batched execution, and episode reset, while a dependency-isolated client/server architecture separates policy inference from environment execution, so that each side retains its native software stack and may run locally or remotely. The ecosystem integrates 42 robot policies and standardizes their installation, debugging, serving, and evaluation workflows. Across these adapters, model-specific code varies by an order of magnitude while the environment-facing loop stays within a few lines of a fixed reference, confirming that the contract confines heterogeneity to the policy side. In a controlled study, conforming to the standard reduces the integration effort of a representative policy from over five hours to two hours, and packaged agent skills reduce it further to thirty minutes. The same adapters serve RoboTwin, RoboDojo simulation, and standardized real-robot evaluation through one interface. XPolicyLab is released as shared infrastructure for reproducible policy comparison and standardized deployment across simulation and physical platforms. Project website: https://xpolicylab.github.io/.

cs.RO↗

arXiv2026

SMTrap: Cost-Effective DoS Attacks Against Large Reasoning Models via SMT Conflict Guidance

Existing LRM-DoS methods rely heavily on model feedback to synthesize attack queries, requiring either repeated queries to the target model or training a dedicated attack model. These expensive operations severely weaken attack leverage. In this paper, we propose \emph{search amplification}, a novel, model-feedback-free LRM-DoS paradigm. It employs the conflict count derived from an Satisfiability Modulo Theories (SMT) solver as a low-cost external signal to guide the synthesis of inference-heavy Constraint Satisfaction Problem (CSP) instances. Our key observation is that LRMs depend on trial-and-backtracking search when solving CSPs, where higher SMT conflict counts on a given CSP instance positively correlate with more extensive LRM backtracking search and substantially longer output trajectories. Building on this finding, we propose \textsc{SMTrap}, a lightweight, CPU-only framework. Guided by SMT conflict counts, \textsc{SMTrap} generates inference-heavy CSP queries without model queries, attack-model training, or GPU computation. Evaluations across seven frontier models demonstrate the state-of-the-art LRM-DoS capability of \textsc{SMTrap}, producing DoS effects multiple times stronger than existing baselines. To mitigate the threat of \textsc{SMTrap}, we demonstrate a tool-based mitigation that significantly cuts token usage.

cs.CL↗

arXiv2026

SAFE-Pruner: Semantic Attention-Guided Future-Aware Token Pruning for Efficient Vision-Language-Action Manipulation

Real-time inference of vision-language-action (VLA) models is essential for robotic control. While visual token pruning has shown strong potential for accelerating inference, most existing methods mainly base pruning decisions on shallow-layer cues and risk discarding visual information required by deep layers. To address this issue, we propose SAFE-Pruner, a plug-and-play pruning framework that incorporates attention cues of future layers into pruning decisions. Specifically, we identify semantic attention consistency, the tendency that VLA models concentrate their attention probability mass on the same semantic entity across control timesteps. Based on this observation, we design a forward-looking strategy to forecast the token saliency in deep layers, which prevents the premature removal of critical tokens and leads to more stable acceleration. We further introduce a reference timestep refresh strategy that triggers updates upon attention shifts, thereby improving forecasting accuracy and pruning reliability. Extensive experiments across diverse evaluation settings demonstrate that our method achieves up to 1.89x speedup with a minimal degradation in success rate of less than 1.5%, while outperforming state-of-the-art methods by up to 1.9%.

cs.CV↗

arXiv2026

Can We Defend Against AI-Generated Video Attacks on Real-World Crisis Events? A Systematic Evaluation of Detectors, Generators and Social Dissemination

Recent video generators can fabricate realistic depictions of wars, disasters, public emergencies, and other real-world crises, creating substantial risks of misinformation. Existing benchmarks, however, provide limited evidence on detector and generator behavior in such settings, including how detectability varies with generation conditions, how people perceive generated videos, and whether detectors remain reliable during social dissemination. To address this gap, we introduce RA-Bench, a benchmark for AI-generated video detection that uses Real videos as Anchors. RA-Bench contains 17,886 videos, comprising 1,830 real-video anchors across 10 social-risk categories and 16,056 generated clips from four open-source and five closed-source generators. Based on RA-Bench, we organize our evaluation along three dimensions. We first assess detector generalization across seven traditional detectors, ten zero-shot multimodal models under three review settings, and two MLLMs specifically fine-tuned on AI-generated video detection. Across these methods, none of the three detector families generalizes consistently across RA-Bench instances. We then examine how detectability varies with generation quality, conditioning information, and sampling seeds. These analyses show that generation properties affect detector families differently, while source-level detection patterns remain stable across seeds. Finally, we study human authenticity judgments and detector reliability during social dissemination. We find that videos that mislead people are also difficult for current detectors, and that social dissemination makes detection harder. Together, these findings show that current methods struggle to detect realistic AI-generated videos, highlighting the need for detectors robust to evolving video generators.

cs.CV↗

arXiv2026

tSymPerturb converts longitudinal symptom networks into time-indexed intervention strategies

Longitudinal symptom networks encode directed prediction across measurement occasions, but outgoing connectivity does not by itself identify which symptom should be modified, how strongly it should be changed, or how a perturbation would propagate to later symptoms. We introduce tSymPerturb, a temporal extension of SymPerturb for cross-lagged panel networks (CLPNs). The framework separates source-state operators (temporal virtual knockout and knockdown), transition operators (directed edge and source-node communication blocking), and strategy procedures (dosage perturbation, combination analysis and sequence optimisation). For a two-wave linear CLPN, the central propagation identity is $Δμ_2 = B(μ_1 - μ_1^*)$, which makes the source time, outcome time and transition operator explicit. The formulation also yields three falsification constraints: dose response is exactly linear under a fixed linear transition model and linear dose map; independent source-state perturbations are additive at the mean level; and genuine treatment order is not identified from a single two-wave transition. In a known 22-node, four-module generating system, analytical temporal-knockout responses agreed with 250,000-draw Monte Carlo estimates within 0.0057 standard deviations. Across 200 independently generated datasets, median Spearman correlation with the population tVPPS ranking increased from 0.76 at n=250 to 0.88 at n=500 and 0.93 at n=1,000; median top-five recovery was 0.60, 0.80 and 0.80, respectively. Multi-wave simulations showed that target profiles can change across propagation horizons despite high overall rank concordance. tSymPerturb therefore converts longitudinal network structure into auditable, time-indexed intervention hypotheses while retaining the distinction between prediction and causal treatment effects.

q-bio.QM↗

arXiv2026

GigaBrain-0.7: Scaling Embodied Foundation Models to Emergent Capabilities with a Three-System Architecture

Vision-language-action (VLA) models have become a dominant paradigm for generalist embodied agents, demonstrating strong complex and long-horizon task completion in structured settings. Yet it remains an open question whether current VLA systems can benefit from more effective architectural design, scale to substantially larger and more heterogeneous data regimes, and achieve broader generalization across tasks and embodiments. To this end, we present GigaBrain-0.7, an embodied foundation model with substantially improved generalization across diverse robot embodiments. Specifically, GigaBrain-0.7 unifies understanding, prediction, and action through a three-system architecture, scales pretraining to over 37,000 hours of heterogeneous embodied data, and introduces one-stage alignment training that jointly optimizes vision-language understanding and multi-embodiment action generation. Compared with the preceding GigaBrain-0 series and prior state-of-the-art models including $π_{0.5}$, GigaBrain-0.7 achieves substantial improvements in foundation zero-shot capabilities, language-conditioned instruction following, and post-training task success rates. In particular, on our in-house Maker H01 platform and mainstream robot embodiments, GigaBrain-0.7 demonstrates strong task adaptability and completion ability across both home and industrial scenarios. All training code and pretrained model weights will be released.

cs.RO↗

arXiv2026

ORV: 4D Occupancy-centric Robot Video Generation

Recent embodied intelligence suffers from data scarcity, while conventional simulators lack visual realism. Controllable video generation is emerging as a promising data engine, yet current action-conditioned methods still fall short: generated videos are limited in fidelity and temporal consistency, poorly aligned with controls, and often constrained to singleview settings. We attribute these issues to the representational gap between sparse control inputs and dense pixel outputs. Thus, we introduce ORV, a 4D occupancy-centric framework for robot video generation that couples action priors with occupancy-derived visual priors. Concretely, we align chunked 7-DoF actions with video latents via an Action-Expert AdaLN modulation, and inject 2D renderings of 4D semantic occupancy into the generation process as soft guidance. Meanwhile, a central obstacle is the lack of occupancy data for embodied scenarios; we therefore curate ORV-Data, a large-scale, high-quality 4D semantic occupancy dataset of robot manipulation. Across BridgeV2, DROID, and RT-1, ORV improves video generation quality and controllability, achieving 18.8% lower FVD than state of the art, +3.5% success rate on visual planning, and +6.4% success rate on policy learning. Beyond singleview generation, ORV natively supports multiview consistent synthesis and enables simulation-to-real transfer despite significant domain gaps. Code, models, and data are at: https://orangesodahub.github.io/ORV

cs.CV↗

arXiv2026

Pair-Density Wave from Doping an Altermagnetic Mott Insulator

Pair-density-wave (PDW) superconductivity is a state in which the superconducting order parameter modulates at a finite wavevector. Using large-scale density matrix renormalization group, we study the doped altermagnetic Mott insulator in the checkerboard $t$-$J$ model, where altermagnetic exchange anisotropy is encoded microscopically through anisotropic ferromagnetic next-nearest-neighbor exchange. By mapping the ground-state phase diagram as a function of doping and altermagnetic anisotropy, mainly on six-leg cylinders, we identify a transition from a uniform $d$-wave superconducting regime with charge modulation to a PDW regime coexisting with stripe order. In the PDW regime, we report an unconventional wave-vector locking $\mathbf Q_{\mathrm{PDW}}\approx 2\mathbf Q_{\mathrm{Stripe}}$ along the cylinder direction, in contrast to the conventional relation. Pair correlations reveal a two-scale structure, consisting of short-distance local $d$-wave pairing and long-distance finite-momentum PDW correlations. A symmetry-based Ginzburg--Landau analysis is presented for the observed locking. Our results identify altermagnetism as a strong-coupling, microscopically grounded route to finite-momentum superconductivity in doped Mott insulators.

cond-mat.str-el↗

arXiv2026

Inverting multiple quantum many-body scars via disorder

The observations of persistent revivals in the Rydberg atom chain have revealed a weak ergodicity-breaking mechanism known as quantum many-body scars, which is typically a collection of states with low entanglement embedded in otherwise thermal spectra. Here, by applying a generic formalism, we reveal a direct evolution from the quantum many-body scars to the multiple inverted quantum many-body scars, i.e., different sets of excited states with volume-law entanglement entropy embedded in a sea of states with area-law entanglement. When increasing the disorder strength, a series of exact eigenstates, acting as conventional QMBS in a regime of weak disorder, remain unchanged. Around each of these states, the inverted quantum many-body scars are introduced by the increased disorder. Moreover, the strong disorder also gives rise to additional sets of inverted quantum many-body scars with their energies concentrating in the middle of the exact eigenstates. As a result, all the multiple inverted quantum many-body scars are approximately equidistant in energy, reminiscent of conventional quantum many-body scarred states. Despite being measure-zero states in the whole spectrum, these inverted quantum many-body scarred states significantly influence nonequilibrium dynamics in the large disorder regime. Random thermal states in a specific subspace show periodic revivals in the fidelity dynamics, while the typical charge-density-wave states exhibit persistent imbalance dynamics. We further examine the stability of the conventional and the inverted quantum many-body scars against the external random field. Our findings expand the variety of nonthermal systems and draw a connection between the weak violation of ergodicity and that of non-ergodicity.

cond-mat.dis-nn↗

arXiv2026

Nonreciprocity reversal of magnetoacoustic attenuation in NiFe alloy thin films

Nonreciprocity, the asymmetry of transport, underlies technologies from the diode to the microwave isolator. In a ferromagnet, a surface acoustic wave generates an elliptical effective field with propagation-locked handedness, breaking the reciprocity of its propagation. Despite decades of study on this phenomenon, a method for controlling the sign of the nonreciprocity has remained elusive. Here we observe a sign reversal in Ni$_x$Fe$_{100-x}$ films. A 0.8 at.% change across Permalloy's zero-magnetostriction composition, where the magnetoelastic coefficient $b$ changes sign, reverses the handedness of the elliptical effective field and thereby the nonreciprocity, from 78.6% to -61.8%. Angle-dependent measurements and spin-wave-ellipticity modelling show that reversing the sign of $b$ reverses the handedness of the elliptically polarized effective field. Aided by cubic frequency scaling, we resolve the sign of $b$ down to -0.05 MPa in a 10-nm film, establishing nonreciprocity as a nanoscale probe of magnetoelastic coupling.

cond-mat.mtrl-sci↗

arXiv2026

Never compromise with vulnerabilities: a comprehensive survey on AI governance

The rapid advancement of AI has expanded its capabilities across domains, yet introduced critical technical vulnerabilities, such as algorithmic bias and adversarial sensitivity, that pose significant societal risks, including misinformation, inequity, security breaches, physical harm, and eroded public trust. These challenges highlight the urgent need for robust AI governance. We propose a comprehensive framework integrating technical and societal dimensions, structured around three interconnected pillars: Intrinsic Security (system reliability), Derivative Security (real-world harm mitigation), and Social Ethics (value alignment and accountability). Uniquely, our approach unifies technical methods, emerging evaluation benchmarks, and policy insights to promote transparency, accountability, and trust in AI systems. Through a systematic review of over 300 studies, we identify three core challenges: (1) the generalization gap, where defenses fail against evolving threats; (2) inadequate evaluation protocols that overlook real-world risks; and (3) fragmented regulations leading to inconsistent oversight. These shortcomings stem from treating governance as an afterthought, rather than a foundational design principle, resulting in reactive, siloed efforts that fail to address the interdependence of technical integrity and societal trust. To overcome this, we present an integrated research agenda that bridges technical rigor with social responsibility. Our framework offers actionable guidance for researchers, engineers, and policymakers to develop AI systems that are not only robust and secure but also ethically aligned and publicly trustworthy. The accompanying repository is available at https://github.com/Tele-EVOL/AI-Governance.

cs.CR↗

arXiv2026

SymPerturb converts symptom-network structure into testable intervention priorities

Symptom networks encode conditional dependence but do not by themselves identify causal or clinically actionable intervention targets. We introduce SymPerturb, a virtual-perturbation framework that distinguishes four primitive perturbation operators - virtual knockout, virtual knockdown, edge-level communication blocking and node-centred communication blocking - from three analytic procedures - virtual dosage perturbation, combination perturbation and sequence optimisation. The reference Gaussian implementation is embedded in a general location-scale map with symptom-specific target anchors, making explicit that zero anchoring and linked mean-variance attenuation are modelling choices. Seven utility outcomes quantify downstream efficacy, dose efficiency, breadth, cross-module reach, communication blocking, combination value and responsiveness; robustness is reported separately as an uncertainty diagnostic. Their direction-aligned, within-candidate-set weighted mean defines the virtual perturbation priority score (VPPS), which is a relative ranking rather than a transportable clinical utility score. In a known 22-node, four-module generating network, analytical efficacy agreed with 100,000-draw Monte Carlo estimates within 0.0024 standard deviations. The reported finite-sample VPPS results were generated with the original eight-component exploratory score and therefore require regeneration under the revised seven-utility-dimension definition. These simulations provide internal computational verification under model compatibility, not causal or external validation. SymPerturb is intended to generate auditable target hypotheses for longitudinal and experimental testing.

q-bio.QM↗

arXiv2026

Zero2Skill: Bootstrapping Robot Skills through Autonomous Data Collection, Training, and Deployment

Autonomous data collection governs the volume and quality of real-world trajectories for manipulation policy learning. Existing pipelines reduce human effort via self-resetting, VLM verification, or language-guided correction, yet episode-scoped fixes must be reissued whenever the same failure recurs, so oversight cost grows with session length rather than with the number of distinct problems. We present Zero2Skill, a human-robot symbiotic agentic system in which corrections are retained and reused across rounds. The collection loop collects, verifies, and resets autonomously, pausing for a remote operator only when a phase exhausts an explicit retry budget. An LLM parser maps each natural-language utterance to a structured adjustment stored in Corrective Memory, so addressed failure modes typically need not be corrected again under the same conditions. On a real-robot desktop-clearing testbed, Zero2Skill matches teleoperation episode success while reducing human working time to 16%. Language corrections improve verifier-human agreement in all four evaluated settings and raise average single-attempt success from 12.5% to 47.5% (arm-selection: 20.0% to 50.0%). Policies fine-tuned on Zero2Skill data match teleoperation-trained policy success at a fraction of collection human cost.

cs.RO↗

arXiv2026

STEREOFLOW: Progressive Stereo Matching with StereoDiT and Transition Flow Matching

Stereo matching is a fundamental task in 3D reconstruction. Despite remarkable advances, the prevailing paradigms formulate stereo matching as a deterministic regression problem, collapsing the multimodal distribution modeling into a single-point estimation. This formulation suffers from a regression-to-mean bias, frequently struggling with ambiguous regions. In contrast, we introduce a prior-guided generative framework that integrates deterministic matching regression and generative distribution modeling within a complementary formulation. Built upon this formulation, we introduce StereoFlow through three key components: (i) a two-stage progressive cascade matching network that progressively produces multi-resolution stereo conditions with complementary matching cues; (ii) a pixel diffusion transformer (termed StereoDiT) with a frequency-decoupled architecture for modeling correspondence ambiguity; (iii) a few-step flow matching objective (termed Transition Flow Matching) for efficient optimization. In summary, \textsc{\textbf{StereoFlow}} achieves strong geometric consistency and rich fine-grained details in ill-posed, discontinuous regions and under zero-shot generalization. Extensive experiments demonstrate that the proposed StereoFlow establishes multiple state-of-the-art results across benchmarks, including Scene Flow, KITTI, ETH3D, and Middlebury.

cs.CV↗

arXiv2026

GigaWorld-Policy-0.5: A Faster and Stronger WAM Empowered by AutoResearch

World Action Models (WAMs) improve robot policy learning by jointly modeling actions and future visual observations, using future scene evolution as dense supervision for physically grounded action generation. However, a common design in existing WAMs is to explicitly generate future videos at inference time, incurring substantial computational overhead and hindering real-time closed-loop deployment. GigaWorld-Policy addresses this issue with an action-centered formulation, where future visual dynamics are used during training while action-only decoding is used at inference time. Building upon this framework, we present GigaWorld-Policy-0.5, an enhanced action-centered WAM designed for more efficient robot control. During pretraining, GigaWorld-Policy-0.5 adopts a mixed Action-Conditioned World Modeling (AC-WM) and WAM training strategy. This strengthens the coupling between visual dynamics and robot actions and improves the transferability of action representations for downstream policy learning. For efficient inference, GigaWorld-Policy-0.5 introduces a Mixture-of-Transformers architecture that separates visual dynamics modeling and action generation into specialized experts, reducing active computation during action-only inference and achieving 85 ms inference latency on a local RTX 4090 setup. In addition, we employ an agent-based AutoResearch pipeline to systematically search training configurations, enabling more efficient identification of optimal experimental setups while reducing the time and manual intervention required for hyperparameter tuning. Experiments and ablations show that GigaWorld-Policy-0.5 preserves the training benefits of future visual dynamics while improving inference efficiency for robot control.

cs.RO↗