Search arXiv⌕ Search

arXiv subjects

Xiaofeng Wang

Publications and source records attributed to Xiaofeng Wang.

At least 19 recordsLinked to original sources
arXiv2026

V$^3$AST-Bench: Verifiable, Valid, and Versatile Safety Evaluation for Coding Agents

Large language models increasingly act through code, modifying repositories, executing programs, and interacting with services in agent form. Their actions introduce risks that cannot be assessed from generated responses alone. Safety evaluation must verify actual consequences, distinguish harm avoidance from task incapability, and cover realistic workflows. We introduce V$^3$AST-Bench, a benchmark for Verifiable results, Valid interpretation, and Versatile risk coverage of Agent Safety. It comprises 212 risk tasks spanning 18 categories. Dedicated runtime environments verify harm and task outcomes through executable checks over states, records, and artifact behavior, while paired capability twin tasks verify workflow competence. Our experiments span 51 combinations of 11 models and five harnesses, spanning 86,496 trials and over 9 billion tokens of model interaction. Results show that V$^3$AST-Bench remains challenging even for frontier coding agents, as Claude Opus 4.8 with Claude Code still causes verified harm on 20.9% of tasks. Across model-harness combinations, agents with comparable competence exhibit markedly different safety outcomes, showing that task capability alone does not guarantee safe execution. More importantly, even explicit recognition of risks does not reliably prevent harmful actions, revealing a critical gap between risk awareness and safe execution. These findings highlight the need to move beyond risk recognition toward enforcing safety throughout execution.

cs.SE↗
arXiv2026

SEMPRE: Semantic-Aware Protocol Reverse Engineering via Consistency-Driven Joint Optimization

Protocol reverse engineering (PRE) is essential for understanding undocumented or proprietary communication formats, enabling security-critical tasks such as vulnerability discovery, malware C2 analysis, and IDS signature generation. However, existing PRE techniques rely on brittle hand-written rules, under-model contextual dependencies among neighboring fields, and often fail on messages with nested sub-protocols. We present SEMPRE, a fully automated PRE framework that replaces ad-hoc heuristics with recursive LLM reasoning grounded in fine-grained dynamic program analysis. SEMPRE uses taint-flow and loop traces to guide an LLM-driven tree-of-thought search that slices messages into fields while detecting nested sub-payloads, producing a hierarchical layout. It then infers each field's type and role and iteratively refines boundaries by freezing high-confidence segments and re-running detection under semantic constraints, yielding an EM-style procedure that monotonically improves structure and semantics until convergence. Evaluated on a broad corpus of mainstream protocols against six state-of-the-art (vs. <0.4 for all baselines) and 0.72 for semantic inference (vs. 0.13 for BinPRE). Ablations confirm the contribution of each component, and a case study on a prolific real-world botnet demonstrates practical utility.

cs.CR↗
arXiv2026

EMMA: Enhancing Robotic Control Robustness through a Generative Visual Model

The generalization of vision-language-action (VLA) models heavily relies on diverse training data. However, acquiring large-scale robot manipulation data across varied object appearances is costly and labor-intensive. To address this, we introduce Embodied Manipulation Media Adaptation (EMMA), a framework for augmenting VLA policies that combines a generative data engine with an effective training pipeline. We introduce DreamTransfer, a diffusion Transformer-based architecture for generating multi-view consistent and geometrically plausible embodied manipulation videos. It supports text-driven video editing on foreground, background and lighting while preserving 3D structure and geometric fidelity. We also utilize a hybrid training set of real and generated data and propose AdaMix to enhance the training process. AdaMix is a training strategy that adaptively weights samples according to policy performance to emphasize challenging samples. Evaluations show DreamTransfer outperforms previous video generation models in multi-view consistency, geometric fidelity, and text-conditioning precision. In real-world robotic tasks with zero-shot visual settings, our framework achieves a relative performance increase of over 92% compared to training with real data alone, and improves by an additional 17% with AdaMix.

cs.AI↗
arXiv2026

State World Models: Execute the Rules, Generate the Pixels

Existing world models can generate action-conditioned video from a single image. Many focus on camera movement, allowing users to explore the generated scene. However, navigation alone does not ensure that an action follows persistent rules when its outcome depends on state that is invisible in the current frame. We present Stateful Harness, a framework that separates world evolution from video rendering to support stateful and rule-governed interaction. It maintains an evolving world state under specified rules, renders each resulting event with a frozen video model, and uses the generated clip to update the world state for the next turn. At each turn, the harness resolves the action selected by the user and compiles the resulting event into an instruction, while a preview clip from an editable 3D scene provides spatial and camera guidance. To evaluate rule compliance, we build WorldState-Bench, a benchmark of 50 counterfactual twin pairs that share an initial frame and action but differ in hidden state and required outcome. Under rule-only prompting, the best of the 11 evaluated video backends produces the required endings for both cases in 32% of pairs. For MiniMax-H3, specifying the required endings raises this rate from 32% to 64%, while the full harness reaches 92% with the same frozen model. These results support separating rule execution from video rendering for world interactions with hidden state.

cs.CV↗
arXiv2026

UDL-RT: A unified differentiable all-sky microwave radiative transfer model with learnable parameterizations

Radiance-based retrievals and data assimilation require radiative transfer models (RTMs) to simulate radiances and quantify their sensitivities to atmospheric and surface conditions. Conventional RTMs often rely on separately maintained forward, tangent-linear, and adjoint codes, complicating model development and extension. Here, we present the Unified Differentiable and Learnable Radiative Transfer model (UDL-RT), a PyTorch-based framework for all-sky microwave radiative transfer that couples neural-network (NN) parameterizations of gaseous absorption, hydrometeor bulk scattering properties (BSPs), and ocean surface emissivity with an explicit radiative transfer solver in a unified differentiable computational graph. Automatic differentiation directly computes Jacobians with respect to input variables, reducing the model maintenance burden. The NN parameterizations offer flexible representations of complex nonlinear relationships. For example, the NN gas-absorption parameterization reduces the standard deviation of brightness-temperature (TB) errors by approximately 60-85% in the water-vapour channels compared with linear regression. Moreover, UDL-RT propagates loss gradients to trainable parameters to optimize physical parameterizations against observations, with GPU-accelerated batched computation supporting efficient iterative learning. Using Microwave Humidity Sounder-2 (MWHS-2) observations, we optimize the snow BSP parameterization through gradient-based learning. The optimized parameterization improves agreement between simulated and observed TB distributions and reduces the systematic bias and skewness of observation-minus-background (O-B) departures. Overall, UDL-RT integrates radiance simulation, sensitivity analysis, and parameterization optimization within a unified framework, supporting future remote sensing retrieval and data assimilation applications.

physics.ao-ph↗
arXiv2026

Fast X-ray Transient EP260321a: Shock Breakout through an Extended Circumstellar Matter

The Einstein Probe (EP) has discovered a fast X-ray transient, EP260321a, whose optical counterpart was subsequently classified as a broad-lined Type Ic supernova. Its observed rise time of about 690 seconds is inconsistent with theoretical predictions for shock breakout (SBO) from a Wolf-Rayet star, which is typically regarded as the progenitor of envelope-stripped supernovae. In this paper, we propose that EP260321a arose from a SBO within extended circumstellar matter (CSM), where breakout photons were initially trapped, then diffused out of the unshocked CSM as the ejecta expanded and the opacity decreased. We modeled this photon diffusion process using both Monte Carlo simulations and an analytical approach to describe the temperature evolution. Through Markov Chain Monte Carlo sampling, we obtain the best-fit CSM parameters of log10(Mcsm/Msun) = -3.3 for the total CSM mass, Rin = 160 Rsun for the inner CSM radius, and Rout = 612 Rsun for the outer CSM radius. The presence of such a dense CSM indicates that the progenitor of EP260321a underwent episodic mass loss just days before the explosion, providing new observational constraints on the late-stage evolution of massive stars.

astro-ph.HE↗
arXiv2026

Training-Data Extraction from Genomic Language Models

Genomic language models generate long nucleotide continuations from short prefixes, exposing a direct route for studying training-sequence extraction through the generation interface. We present an output-only black-box attack on Evo 2 that selects convergent groups of repeated generations without gradients, token likelihoods, or training-corpus access during candidate selection. Recovered content is then scored by an evaluator combining aligned length, nucleotide identity, match multiplicity, and sequence complexity. Across 304 runs on Evo 2 7B with bacteriophage and plasmid prefixes, the attack yields 888 candidate clusters, of which 85 match training segments of at least 500 bp, giving a cluster-level attack success rate of 9.57%. Final filtering retains 22 uniquely matched extraction events at the 500-bp threshold and 15 at 1,000 bp, covering 14 and 7 distinct training-sequence regions, respectively. Within-category source-homology screening retains 11 of the 22 events at a 99% exclusion threshold and four at 98%, with the latter mapping to two reference regions. The results establish convergence among repeated nucleotide generations as a practical output-only signal for recovering long training-sequence segments and characterize how recovery depends on prefix length, cluster size, perplexity, early stopping, and sequence domains.

cs.CR↗
arXiv2026

RobotAPO: Adversarial Physics Preference Optimization for Robotic Manipulation Video Generation

Robotic manipulation videos are increasingly used as visual plans for embodied agents, but optimizing purely for visual plausibility often fails to capture the fragile physical manifold of real-world interactions. Even minor physics-violating errors at the interaction boundary, such as interpenetration or premature object motion, can completely invalidate the inferred timing and pose needed for downstream execution. Because standard supervised fine-tuning lacks the direct pressure to penalize these localized failures, we introduce AgiBot-PhysPref. This rigorously curated 10,000-sample preference dataset isolates condition-matched physics violations, turning the generator's own failure distribution into a foundational signal for physical consistency. Building upon this, we propose RobotAPO, an adversarial physics preference optimization framework operating in the continuous flow-matching denoising space. To prevent the policy from merely memorizing static curated failures, RobotAPO employs a lightweight adversarial counterfactual proposer that learns a condition-dependent, physical-failure-biased direction in denoising space. This encourages the model to explore and better respect the physical interaction boundary, all while maintaining a pure prompt-and-reference inference interface without requiring external structural conditioning. Comprehensive evaluations demonstrate that explicitly correcting these localized physics violations improves downstream robot execution from generated videos. On held-out AgiBot conditions, RobotAPO outperforms the strongest controlled internal baseline in physical consistency by 6.8% hard score and 10.0% soft score. Crucially, in real-robot replay, it translates these physical-consistency gains into a 37.4% relative improvement in task success over the strongest controlled internal baseline.

cs.RO↗
arXiv2026

LEO-based Carrier-Phase Positioning for 6G: Design Insights and Comparison with GNSS

The integration of non-terrestrial networks (NTN) into 5G new radio (NR) enables a new class of positioning capabilities based on cellular signals transmitted by Low-Earth Orbit (LEO) satellites. In this paper, we investigate joint delay-and-carrier-phase positioning for LEO-based NR-NTN systems and provide a convergence-centric comparison with Global Navigation Satellite Systems (GNSS). We show that the rapid orbital motion of LEO satellites induces strong temporal and geometric diversity across observation epochs, thereby improving the conditioning of multi-epoch carrier-phase models and enabling significantly faster integer-ambiguity convergence. To enable robust carrier-phase tracking under intermittent positioning reference signal (PRS) transmissions, we propose a dual-waveform design that combines wideband PRS for delay estimation with a continuous narrowband carrier for phase tracking. Using a realistic simulation framework incorporating LEO orbit dynamics, we demonstrate that LEO-based joint delay-and-carrier-phase positioning achieves cm-level accuracy with convergence times on the order of a few seconds, whereas GNSS remains limited to meter-level accuracy over comparable short observation windows. These results establish LEO-based cellular positioning as a strong complement and potential alternative to GNSS for high-accuracy positioning, navigation, and timing (PNT) services in future wireless networks.

cs.IT↗
arXiv2026

GEM-Occ: From Visual Geometry Evidence to Embodied Semantic Occupancy Memory

Embodied agents exploring indoor environments require reliable semantic occupancy memory that persists across observations and revisits. Building such memory is challenging because each observation provides incomplete and uncertain geometric and semantic evidence. We introduce GEM-Occ, a Gaussian Evidence Memory framework that consolidates evidence accumulated over time into persistent semantic occupancy memory. Local predictions are converted into occupied semantic Gaussians and free-space ray evidence. Confidence- and visibility-aware causal updates integrate supporting observations, suppress occupancy contradicted by observed free space, and preserve previously observed structures through occlusion. A hierarchical memory organization supports continued mapping and efficient queries across connected indoor spaces. To evaluate this capability, we introduce HIOcc, a unified benchmark for embodied semantic occupancy memory. HIOcc establishes a shared semantic label space and evaluation framework spanning local prediction, room-level online mapping, and building-level mapping, while accommodating perspective and panoramic observations. Experiments on HIOcc demonstrate that GEM-Occ outperforms existing methods, enabling accurate semantic occupancy prediction and consistent online mapping across spatial scales with efficient memory usage and fast occupancy queries.

cs.RO↗
arXiv2026

JWST Nebular Spectroscopy of SN 2023qov: Circumstellar Dust Emission in a Normal Type Ia Supernova

We present panchromatic observations of the Type Ia supernova (SN Ia) 2023qov, ranging from $\sim$2 weeks before to $\sim$1 year after maximum light. JWST near- and mid-infrared spectra at $+$276 and $+$363~days show $\sim$400 K dust emission that cools by $\sim$75 K between epochs, the first unambiguous spectroscopic detection of dust emission in a normal SN Ia. We find that the emission is well described by models of carbonaceous dust placed within $\sim$1 light year of the SN, with a dust mass of $\sim$$10^{-4}$ M$_{\odot}$. We do not see evidence of active dust creation, suggesting an infrared light echo by pre-existing circumstellar dust as the likely source of the emission. The JWST nebular line profiles suggest asymmetric, stratified ejecta, similar to other normal SNe Ia, though a slight double-horn structure in the argon lines indicate a toroidal enhancement. SN 2023qov exhibits a slightly red, fast-declining early light curve ($Δm_{15}(B) = 1.47 \pm 0.05$ mag), from which we determine a $^{56}$Ni mass of $M_{56} = 0.21 \pm 0.04$ M$_{\odot}$, and a distance of $d = 36.0 \pm 1.8$ Mpc to the SN and its host, NGC 7029.

astro-ph.HE↗
arXiv2026

SimFuse3D: Source-Guided Target Simulation and Confidence-Guided Multi-Stage Localization Reweighting for Cross-Platform 3D Object Detection

Changes in sensor height and viewpoint alter object-level point distributions, making cross-platform LiDAR unsupervised domain adaptation (UDA) difficult. Self-training uses labeled source scans and unlabeled target scans, yet a retained prediction may provide a useful target location while enclosing sparse foreground returns, background clutter, or points inconsistent with the predicted box. We refer to this mismatch as box-point inconsistency. We introduce SimFuse3D, which preserves the target placement and repairs the associated pseudo object using measured geometry from labeled source scans. Object Memory retrieves a similar labeled source instance. Target Simulation places the retrieved source geometry at the target location, aligns its points with the target viewing geometry, and filters the aligned crop to approximate the target observation. Confidence-Guided Multi-Stage Localization Reweighting (CMLR) maps each target pseudo-object confidence score to a bounded weight shared by RPN localization and R-CNN box regression. All components operate only during adaptation, leaving the detector architecture and inference graph unchanged. Across six cross-platform transfers, SimFuse3D consistently outperforms Pi3DET-Net and achieves the best performance among the compared adaptation methods on nearly all metrics. On nuScenes-to-KITTI, it ranks first among the compared adaptation methods with both evaluated detectors.

cs.CV↗
arXiv2026

Effective Contact Theory for Exotic Loosely Bound States in Strongly Interacting Expanding Matter

Short-distance correlations generate universal relations that can be independent of microscopic details. Tan's contact is a prominent example, connecting close pairs, high-momentum constituents, and bound-state formation across atomic, condensed-matter, and nuclear systems. Whether an analogous universality governs the nonequilibrium freeze-out of relativistic QCD matter is an open question. We develop an effective contact theory that relates the production of loosely bound states to continuum two-particle correlations in heavy-ion collisions. The construction replaces the hard relative-momentum cutoff of conventional coalescence by a regulated Bethe--Peierls kernel: freeze-out localization fixes its momentum scale, while the composite yield fixes its contact residue. Bound and continuum observables then become projections of a common two-particle density matrix rather than independent phenomena. Coulomb-bound $Kμ$ atoms provide an ideal realization because their atomic, production, and source scales are widely separated. A contact-theory analysis of STAR $dΛ$ correlations supplies a first bound--continuum consistency test and demonstrates how inferred near-threshold parameters can depend on the finite expanding source. The framework establishes a general strategy for relating coalescence, correlations, and exotic bound states in strongly interacting expanding matter.

nucl-th↗
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

Closed Range and Essential Norms of Composition Operators on Weighted Dirichlet Spaces

We study closed range and essential norms of bounded composition operators $C_φ$ on weighted Dirichlet spaces $\mathcal{D}_α$. For $-1<α<0$, we first consider maps whose images are obtained by removing a compact subset from a simply connected subdomain of $\mathbb{D}$. In this setting, closed range is equivalent to the Reverse Carleson property and the Reverse Carleson Disk Condition for the counting measure $μ_{φ,α}$, uniform weighted area density of the image, $\mathbb{T}\subset\overline{φ(\mathbb{D})}$, and the inclusion of an outer annulus in the image. We also obtain a closed range characterization under the Uniform Tail Condition. For all $α>-1$, we establish two-sided essential norm estimates in terms of local averages of the generalized Nevanlinna counting function and an exact formula using boundary tail integrals. If the normalized counting density has vanishing oscillation at the boundary, we obtain exact formulas in terms of its boundary limsup, Berezin transform, and local averages.

math.FA↗
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

Towards TEE-Certified DP: Verifiable Differentially Private Training on Legacy GPUs

Wide adoption of machine learning has created growing policy and regulatory demand for protecting sensitive training data, with differential privacy (DP) emerging as a key mechanism. Yet a less-studied problem is how to certify the faithful execution of DP during training: an external verifier should be able to check that a released model was trained with proper DP protection, without accessing the private training data. Existing cryptographic approaches, such as zero-knowledge proofs, provide strong guarantees but often incur prohibitive overhead, in some cases by orders of magnitude. Trusted Execution Environments (TEEs) offer a more efficient alternative, but the multi-GPU TEE support needed for training and fine-tuning large language models remains limited to recent platforms and is absent or inefficient on legacy GPUs. To address this, we propose a practical framework for verifiable DP training using CPU-side TEEs together with untrusted GPUs. Our design addresses a fundamental efficiency-security tension: training entirely inside a CPU TEE is too slow, while unrestricted GPU offloading can allow malicious deviations from DP. We therefore offload expensive gradient computation to GPUs, while using the CPU TEE to efficiently verify the correct enforcement of DP on gradients through probabilistic checking. Our framework detects frequent full deviations from DP with high probability; for the utility-oriented forged-gradient attacks evaluated in this work, sparse deviations provide limited utility benefit and show no measurable additional membership leakage. Experiments further show that our approach nearly achieves a ``free lunch'': it incurs only modest overhead compared with standard GPU-based DP training, while effectively constraining malicious deviations from the claimed DP execution.

cs.CR↗
arXiv2026

Bergman Kernels and Hankel Operators on Weighted Fock Spaces in Several Complex Variables

We introduce a class of plurisubharmonic weights on \mathbb C^n and develop a theory of the associated weighted Fock spaces F_φ^p. The curvature assumptions are imposed on a \mathcal C^2-regularization at bounded distance from the original weight. Thus the original weight need not be smooth or strictly plurisubharmonic, and, even when it is of class \mathcal C^2, its complex Hessian eigenvalues need not be uniformly comparable. This provides a counterpart of Christ's doubling theory in several complex variables. We establish a global upper estimate with decay away from the diagonal and a uniform lower estimate near the diagonal for the weighted Bergman kernel. These estimates yield L^p bounds for the kernel functions, boundedness of the Bergman projection, and duality and complex interpolation for the associated Fock spaces. For a certain class of weights beyond the uniformly controlled curvature setting, we construct an integral solution operator for the \bar\partial equation and establish scale-adapted weighted L^p estimates for 1\leq p\leq\infty. As an application, for all 1\leq p,q<\infty, we characterize the boundedness and compactness of Hankel operators from F_φ^p to L_φ^q with possibly unbounded symbols in terms of weighted IDA spaces.

math.CV↗