Search arXiv⌕ Search

arXiv subjects

Hao Wu

Publications and source records attributed to Hao Wu.

At least 127 records · Page 7Linked to original sources
arXiv2026

Astrophysics on GPUs: introducing AGILE 1.0

We present AGILE, a GPU-enabled adaptive mesh refinement (AMR) framework for the solution of (near-) conservation laws which occur in astro- and solar-physical applications. AGILE is written in modern fortran 2003, inherits a part of its modules and mesh handling from MPI-AMRVAC, and achieves excellent GPU performance via OpenACC offloading. We here discuss the design decisions which enable AGILE to perform cost-efficient and scalable deeply nested AMR simulations with moderate block sizes of e.g. $16^3$ cells. AGILE currently implements several physics modules, ie. hydrodynamics, frozen-field hydrodynamics, magnetohydrodynamics and special-relativistic hydrodynamics and can easily be extended further through its modular design. Besides strong scaling tests to up to 2048 GPUs and standard benchmarks which show consistent performance across a large range of devices and problem sizes, we demonstrate AGILE's capabilities by means of state-of-the art science applications with all currently available physics modules.

astro-ph.IM↗
arXiv2026

TrapHunter: Exposing Covert Pathways in Trap Token Contracts

Standardized token contracts (e.g., ERC-20) form the foundation of digital assets. However, attackers increasingly abuse this standardization to disguise malicious trap tokens. Unlike obvious violations, these contracts employ a strategy of "deceptive adherence": they strictly adhere to standard protocols to evade detection while embedding covert logic to defraud users. To address this, we first systematize the trap landscape by proposing a novel taxonomy derived from the intrinsic functional lifecycle of tokens (Generation, Circulation, Persistence, and Observation). We then propose TrapHunter, a framework designed to identify these traps and expose covert pathways within these deceptive contracts via intent deviation analysis. Specifically, TrapHunter introduces a unified semantic representation combining Abstract Behavior Trees (ABTs) and Augmented Path Graphs (APGs) to normalize intra-procedural syntax and reveal hidden execution paths driven by inter-procedural state dependencies. Crucially, it bridges the semantic gap by leveraging LLMs to reason about the behavioral intent of deviations from reference implementations, followed by fork-based dynamic validation to confirm exploitability. Experimental evaluation on 269 real-world contracts with three LLMs (DeepSeek, GPT, and Gemini) demonstrates that TrapHunter effectively detects all six categories of traps, achieving an average precision of 81.8% and recall of 85.4%, significantly outperforming state-of-the-art tools.

cs.SE↗
arXiv2026

Electrostatic-Elastic Softening and Ultraviolet Instability Driven by Non-DLVO Interactions in Charged Colloidal Crystals

Colloidal crystals permeated by mobile ions exhibit a coupling between electrostatic and elastic degrees of freedom that renormalizes the effective screening length and induces wave-vector-dependent elastic softening. Building on our recently proposed continuum model, we perform a rigorous Gaussian fluctuation analysis to elucidate the stability limits of the homogeneous phase. By integrating out the electrostatic fluctuations, we derive the effective elastic modulus $Γ(q)$ as a function of wave vector $q$. We show that the modulus in the long-wavelength limit ($q \to 0$) remains identically equal to a bare modulus protected by perfect ionic screening. In contrast, the modulus in the short-wavelength limit ($q \to \infty$) softens as the electrostatic-elastic coupling strength $ξ$ increases, vanishing at a critical value $ξ=1$. For $ξ>1$, the fluctuation spectrum exhibits a negative eigenvalue for all wave vectors $q$ larger than a critical (effective screening) wave vector $q_c$, signaling an ultraviolet instability of the uniform phase. In a real colloidal crystal, this divergence is regulated by the discrete lattice cutoff $q_{\max}\simπ/a$, confining the physical instability to a finite band $q_c < q < q_{\max}$. The macroscopic limit $q\to 0$ remains unconditionally stable for all $ξ$. The transition at $ξ=1$ thus marks the onset of short-wavelength mechanical failure, while macroscopic elastic stiffness remains intact. Our analysis clarifies the proper physical interpretation of the minimal coupling model and provides a consistent picture of how non-DLVO interactions can drive local structural collapse in charged colloidal crystals.

cond-mat.soft↗
arXiv2026

Thinking in Video: Can Video Generators Really Reason About the Real World?

Recent advances in world models and video generation have given rise to an emerging reasoning paradigm that leverages video generative models to simulate, predict, and reason about real-world dynamics. We redefine this paradigm as Thinking in Video, where video is not merely an output artifact but a medium for constructing, extending, and verifying causal thought. However, this promise remains unverified: convincing rollouts may reflect memorized appearances rather than causal understanding, while existing metrics separate perceptual fidelity from semantic logic. To evaluate whether video generators support such reasoning, we introduce the Causal-Generative Dual-Judge (CGDJ), auditing World Model Consistency from two perspectives. Explicit Causal Perception tests whether a generator reads a video scenario as a reasoning problem through spatio-temporal flattened visual question answering, while Implicit Generative Perception-Prediction Gap evaluates whether it renders the causal consequence as a consistent future video. Applying CGDJ to representative open- and closed-source generators reveals a clear Perception-Prediction Gap: open-source models produce plausible dynamics despite near-zero explicit causal perception, whereas advanced closed-source systems show stronger but still limited alignment between reasoning and generation. Further analysis exposes audio-visual misalignment, where models verbalize correct causal logic more reliably than they render it, challenging the "world simulator" narrative.

cs.CV↗
arXiv2026

Mixing Configurations for Downstream Prediction

Clustering-based features are widely used in machine learning, but most methods must choose a resolution -- a choice that is global, fixed, and ad hoc. Recent work shows that varying the resolution parameter produces only a finite set of structurally stable partitions, known as configurations. Based on this, we introduce Configuration-Mixed Prediction (CMP), a setting where models learn to adaptively weight these configurations per sample for downstream prediction. We propose MixConfig, a plug-and-play feature augmentation module that extracts configurations from any frozen embedding and learns energy-aware mixing weights via a novel selector that jointly reasons about sample context, cluster assignments, and stability statistics. Experiments across tabular, molecular, vision, and text domains demonstrate consistent improvements over single-resolution and static baselines across diverse predictor architectures, with gains particularly pronounced in low-data regimes.

cs.LG↗
arXiv2026

Machine Learning for Electrode Materials: Property Prediction via Composition

In this work, we benchmark three leading composition based Machine Learning (ML) frameworks, MODNet, CrabNet, and a random forest model based on Magpie features, predicting the properties of battery electrode materials using the Materials Project Battery Explorer dataset. We evaluate these models based on predictive accuracy, visualize numerical features using two-dimensional embeddings, and quantify performance using standard metrics. Our results demonstrate that CrabNet consistently outperforms the other models across all tests. To validate these findings, we employ bootstrap resampling and two cross-validation (CV) strategies (leave-one-cluster-out and stratified 5-fold CV), comparing each model against a control baseline, using unseen experimental data as a hold-out test. We also apply unsupervised clustering using t-SNE and DBSCAN on physically observed features extracted from matminer, revealing coherent material groupings without prior labels. The final selected model consistently improves over controls, and we believe can be used as a early stage oracle for electrode materials composition screening. Our study aims to identify the error distributions and limitations of the approach, discussing the challenges with developing robust ML models. Despite these constraints, our findings suggest the final selected model is effective for early-stage compositional screening.

cond-mat.mtrl-sci↗
arXiv2026

SciForge: An AI-Native, Multimodal Workbench for Scientific Discovery

Scientific work increasingly spans heterogeneous artifacts -- papers, code, datasets, scientific file formats, model outputs, figures, manuscripts, and team decisions -- yet general-purpose AI assistants rarely preserve these objects as a coherent, auditable research state. We present SciForge, a multimodal research-native AI workbench that reserves the graphical interface for human judgment while search, parsing, model routing, workflow execution, plotting, writing, and presentation generation run as modular agent-accessible services. SciForge is built around five pillars: (i) \emph{goal-scoped scientific decision governance} for \textbf{goal-oriented} research, with review gates and shared review surfaces; (ii) \emph{translate-then-reason} for \textbf{multimodal} input, routing scientific objects through domain translators before the agent reasons; (iii) \emph{evidence governance} for \textbf{auditable} traceability, linking claims to provenance chains and audit findings; (iv) \emph{collaborative team science} for \textbf{collaborative} research, enabling multi-role decision governance, with shared team workspaces planned for future releases; and (v) \emph{real-world application scenarios} for \textbf{practical} impact, demonstrated through eight end-to-end user cases, with flagship demonstrations including multi-day agentic research sprints for gene discovery, AI-guided de novo protein design, molecular optimization, and genome-to-BGC discovery. The system combines a thin interaction layer, contextual research capability patterns, an Agent Runtime and Workflow Engine, an Evidence-DAG audit sidecar and a Scientific Model Router. SciForge currently runs as a desktop application, with mobile supervision support; future releases will deepen team collaboration. The system is open-source and available at https://github.com/AGI4Sci/SciForge

cs.AI↗
arXiv2026

Diarization-Guided Qwen-ASR Adaptation for Multilingual Two-Speaker Conversational Speech

This paper describes our self-designed system for Task 1 of the MLC-SLM 2026 Challenge for multilingual two-speaker conversational speech. The system combines a modular speaker diarization front end with a challenge-adapted Qwen3-ASR-1.7B recognizer. The diarization front end performs voice activity detection, subsegment generation, CAMPPlus speaker embedding extraction, two-speaker spectral clustering, and RTTM-based audio segmentation. The resulting speaker-attributed segments are grouped by language or region and decoded by the adapted ASR model. For ASR adaptation, we first perform supervised full fine-tuning on the official training data, then apply LoRA fine-tuning with synthetic speech generated by a three-pipeline TTS-based synthetic speech augmentation framework, and finally refine the model using GRPO reinforcement learning with rewards based on WER/CER and penalties for hallucination, repetition, and length deviation. On the official development set, the full system achieves an average tcpMER of 23.70, reducing the error rate by 6.83 absolute points relative to the released Qwen-ASR-1.7B performance. On the final evaluation set, the system achieves an average tcpMER of 17.97. Ablation results show that supervised fine-tuning provides the largest gain, while synthetic-speech LoRA adaptation and reinforcement learning further improve robustness.

cs.CL↗
arXiv2026

Entropy-Driven Initiation and Cellular Uptake Mediated by Viscoelastic Cytoskeleton: A Kinetic Phase Diagram from Onsager Variational Principle

A fundamental question in receptor-mediated endocytosis remains unanswered: what initial driving force brings ligands and receptors into close proximity? While previous models assume pre-existing contact and overlook this initiation problem, we propose that entropic forces from nanoscale biomolecules in crowded cellular environments provide the essential driving mechanism. We develop a unified continuum model rooted in the Onsager variational principle, where engulfment depth serves as the generalized coordinate and the driving force derives from a free energy landscape of entropic, binding, membrane, and cytoskeleton contributions. The framework naturally incorporates: (i) entropy-driven adhesion as initiation; (ii) ligand-receptor binding as the sustaining force; (iii) membrane deformation via the Helfrich-Canham Hamiltonian; and (iv) cytoskeleton viscoelasticity through the elastic-viscoelastic correspondence principle. The kinetic phase diagram predicts a critical biomolecule concentration for initiation, a lower bound of ligand density for complete engulfment, a finite size window for engulfable particles, and an optimal virus radius of 30--60 nm that decreases with increasing binding energy. The Onsager solubility condition naturally yields the phase boundaries. The model exhibits asymptotic consistency with the classic Asakura-Oosawa result in the large-particle flat-surface limit. Stiffer cells lead to longer engulfment times and narrower size windows. Strikingly, the optimal size matches HIV-1 dimensions under physiologically realistic parameters. This work provides a variational foundation for cellular uptake with implications for virology, nanotechnology, and drug delivery.

cond-mat.soft↗
arXiv2026

Mind the Gap: Action Rebinding Attacks against Android GUI Agents

Large multimodal model powered GUI agents are emerging as high-privilege operators on mobile platforms, entrusted to perceive screen content and inject inputs across application boundaries. While these agents aim to automate complex tasks, we demonstrate that their design introduces a fundamental conflict with Android's strict application sandboxing. We present a novel cross-application Action Rebinding attack, which allows a malicious application with zero dangerous permissions to hijack the agent's execution and perform privileged operations on behalf of the attacker. Our attack exploits the inevitable observation-action gap inherent in the agent's reasoning pipeline. A malicious app can render a benign ``contextual carrier'' to elicit a planned action, and then swap the foreground to a sensitive target application during the reasoning latency. The agent, unaware of the transition, unwittingly executes the action in the privileged context. We further advance this attack by weaponizing the agent's own task-recovery logic to create programmable, multi-step exploit loops , and introducing an Intent Alignment Strategy (IAS) that manipulates the agent's reasoning to rationalize the hijacked state. We evaluate our attack on six widely-used Android GUI agents. Our results demonstrate a 100% success rate for atomic action hijacking and the ability to orchestrate high-impact exploits, including unauthorized file deletion, SMS transmission, and app uninstallation, without the attacker holding any corresponding permissions. Furthermore, since the malicious application separates intent from capability and contains no privileged API calls, it achieves a 0% detection rate across commercial malware scanners (e.g., VirusTotal), highlighting a critical blind spot in current mobile security analysis. To access experimental logs and demonstration videos, please contact yi_qian@smail.nju.edu.cn.

cs.CR↗
arXiv2026

SOAP, Muon, and Beyond: Pushing LLM Pretraining Scales

Higher-order optimizers such as Muon and SOAP offer faster convergence than AdamW, but their computational cost and numerical stability challenges have limited adoption at scale. In this work, we adapt and enhance preconditioned gradient methods to overcome the practical challenges of large-scale LLM pretraining. We first identify instabilities in SOAP at large batch sizes and propose algorithmic modifications including per-step QR orthogonalization and improved preconditioning strategies that eliminate loss spikes and enable stable training in these regimes. We then present a unified empirical study of SOAP, Muon, and AdamW using update-RMS matching to ensure fair learning rate transfer across optimizers. As part of this analysis, we empirically evaluate the orthogonalization quality of Muon. Our experiments on multi-billion-parameter models trained on trillions of tokens reveal that SOAP and Muon consistently outperform AdamW at the scales we tested. Notably, at batch sizes of up to 100M tokens for next-token prediction, these optimizers maintain training stability and quality while AdamW degrades. To enable efficient training at large scale, we introduce a layer-wise distributed optimizer compatible with Megatron-LM. Our implementation balances memory and hides communication while avoiding approximations to the optimizer computations, thus retaining their convergence benefits. Additionally, we identify and build specific system-level improvements to further accelerate our layer-wise implementation. To support the research community, we release a codebase that contains emerging algorithms for optimization: https://github.com/NVIDIA-NeMo/Emerging-Optimizers

cs.LG↗
arXiv2026

EmbodiSkill: Skill-Aware Reflection for Self-Evolving Embodied Agents

Embodied agents can benefit from skills that guide object search, action execution, and state changes across diverse environments. Since embodied environments vary across layouts, object states, and other execution factors, these skills must self-evolve from trajectories generated during task execution. However, existing skill self-evolution methods are mainly developed in digital environments and often convert trajectories into coarse skill updates. Directly applying this paradigm to embodied settings is problematic, because a failed task execution may reflect not only incorrect skill content, but also an execution lapse in which the agent fails to follow valid guidance. We propose EmbodiSkill, a training-free framework for embodied skill self-evolution through skill-aware reflection and targeted revision. EmbodiSkill interprets each trajectory with respect to the current skill, uses skill-changing evidence to update the skill body, and uses execution-lapse evidence to preserve and emphasize valid guidance. Experiments on ALFWorld and EmbodiedBench show that EmbodiSkill consistently improves embodied task success. On ALFWorld, EmbodiSkill enables a frozen Qwen3.5-27B executor to reach 93.28% task success, outperforming GPT-5.2 used as a direct agent without skills by 31.58%. These results show that skill-aware self-evolution helps embodied agents accumulate reusable procedural knowledge from their own trajectories.

cs.AI↗
arXiv2026

Low-Altitude ISAC With Spherical Directly-Connected Antenna Array: Performance Analysis and Beamforming Optimization

The safety development requirements of low-altitude economy (LAE) renders the robust low-altitude airspace monitoring critical important than ever before. Integrated sensing and communication (ISAC) as one of the key development trends of 6G provides potential solutions for the LAE. However, conventional antenna arrays suffer from limited three-dimensional (3D) sensing coverage and degraded angular resolution at high elevation angles. To address these challenges, this paper investigates low-altitude ISAC systems enabled by the recently proposed spherical directly-connected antenna array (DCAA). By carefully deploying multiple simple uniform planar arrays (sUPAs) over a spherical surface, without relying on any phase shifter, spherical DCAA enjoys advantages of full 3D coverage, superior and uniform angular resolution, enhanced energy-focusing and low hardware cost. In this paper, we first characterizes the sensing performance of the spherical DCAA, in terms of the sensing signal-to-noise ratio (SNR), area average probability of detection, and Cramér-Rao lower bounds (CRLBs) for both elevation and azimuth angle estimation. Then, a low-altitude ISAC optimization problem is formulated to maximize the worst-case sensing SNR over a prescribed aerial region while satisfying the communication quality-of-service requirements of ground users. To effective solve this mixed-integer non-convex problem, we develop a novel greed-based joint array selection and beamforming optimization framework. Simulation results demonstrate that spherical DCAA significantly outperforms conventional UPA in terms of sensing coverage, angle estimation accuracy, and communication-sensing SNR tradeoff, highlighting its potential for future low-altitude ISAC systems.

eess.SP↗
arXiv2026

HARC: Coupling Harmfulness and Refusal Directions for Robust Safety Alignment

Understanding how aligned LLMs internally represent safety is critical for diagnosing alignment vulnerabilities, as it explains why jailbreaks succeed and informs the design of robust alignment strategies. Prior work shows that aligned LLMs encode harmfulness and refusal as separable directions in the residual stream at prompt-side token positions. We show that jailbreaks succeed at prompt encoding by suppressing either the refusal or harmfulness direction before any token is generated, with distinct attack classes occupying separable regions of the harmfulness-refusal plane. Extending the analysis to response-token positions, we find that the model recognizes harmful content while it is generating that content, even when it failed to recognize the input as harmful at the prompt side. Motivated by our findings, we introduce HARC (Harmfulness-And-Refusal Coupling), a fine-tuning method that pairs the two directions across both prompt and response positions. Since the intervention is confined to the harmfulness-refusal subspace, it leaves the rest of the residual stream intact and does not degrade general capability or inflate over-refusal. Across extensive experiments, HARC achieves the strongest robustness-capability-usability trade-off among six baselines spanning the major training-time and inference-time safety methods. The harmfulness and refusal directions at prompt and response positions transfer across the five model families and two scales we tested without architecture-specific tuning.

cs.AI↗
arXiv2026

Green Haven or Risky Venture? Exploring the Connectedness and Hedging of Sustainable Cryptocurrencies and Green Financial Markets

Conventional cryptocurrency often leads to increased energy consumption and carbon emissions, while sustainable cryptocurrencies possess the potential to become a green alternative in portfolio management. This study aims to investigate the time-varying connectedness between sustainable cryptocurrency and green financial markets as well as hedging performance when facing market shocks, including COVID-19 and Russia-Ukraine war. TVP-VAR model with Fourier transform and Multivariate GARCH models are employed. The findings indicate that the pairwise connectedness between the sustainable cryptocurrencies and green financial markets has been at a low level, providing diversification benefits in investment portfolio. Besides, short-term connectedness dominates medium- and long-term connectedness. Sustainable cryptocurrencies show higher hedging effectiveness than traditional cryptocurrency.

stat.AP↗
arXiv2026

Guiding Human Validation of LLM-Generated Code via Verifiable Literate Programming

Vibe coding democratizes software development by allowing users to generate code via natural-language (NL) interaction with large language models (LLMs). However, the code is reliable only when it faithfully implements the user's intent, which is difficult and labor-intensive for users to validate. Existing validation methods either rely on LLM-assisted automated testing, which suffers from prompt ambiguity and model fallibility, or involve users only in partial software artifacts such as prompts and test cases, which may overlook corner cases and program details. Motivated by a bug study of LLM-generated code, we find that detailed human feedback is essential, as failures often stem from underspecified requirements or subtle semantic deviations. This paper presents verifiable literate programming (VLP), a human-in-the-loop framework designed to make the review/validation process of LLM-generated code accessible to users at all programming levels. At its core, VLP proposes unambiguous NL-based documentation as a readable intermediate layer between prompts and code. The documentation demonstrates concrete program semantics and enables users to provide feedback on potential intent-code mismatches. It supports human-involved, end-to-end repair and validation via three techniques: (i) an NL-style literate language with unambiguous syntax and mostly deterministic code-to-documentation translation, (ii) LLM-based fine-grained mismatch detection that uses trace links between prompts and documentation to focus users' review effort on suspicious documentation lines, and (iii) a verification module that leverages user-validated documentation to derive API-usage checks and formal properties, which are then verified against the generated code using model checking. Our evaluation shows that VLP improves code pass@1 from 28.7%-73.2% to 65.4%-93.5% with reasonable user effort.

cs.SE↗
arXiv2026

Latent Visual Cache for Video Reasoning

Video reasoning requires Large Multimodal Models (LMMs) to remain grounded in dense evidence, yet existing systems largely adopt "read-once, generate-many" paradigm, in which visual grounding weakens during generation. This phenomenon has been widely observed and is known as Visual Anchoring Decay. To fill this gap, we introduce Latent Video Cache (Latent-VC), a recurrent latent visual cache inserted into the decoder to preserve compact visual memories throughout reasoning. The cache is trained with supervised contrastive cache alignment and vision-grounded GRPO with a latent grounding reward, while maintaining strict train-inference alignment through native decoder hidden states. Built on Qwen3.5-9B, Latent-VC consistently outperforms strong CoT and SFT+GRPO baselines across six video benchmarks, with especially clear gains on grounding-intensive and long-video tasks. In addition, it also achieves higher accuracy with substantially shorter responses, suggesting that latent visual caching improves video reasoning by preserving visual evidence rather than relying on longer textual chains.

cs.CV↗
arXiv2026

Em-Garde: A Propose-Match Framework for Proactive Streaming Video Understanding

Recent advances in Streaming Video Understanding has enabled a new interaction paradigm where models respond proactively to user queries. Current proactive VideoLLMs rely on per-frame triggering decision making, which suffers from an efficiency-accuracy dilemma. We propose Em-Garde, a novel framework that decouples semantic understanding from streaming perception. At query time, the Instruction-Guided Proposal Parser transforms user queries into structured, perceptually grounded visual proposals; during streaming, a Lightweight Proposal Matching Module performs efficient embedding-based matching to trigger responses. Experiments on StreamingBench and OVO-Bench demonstrate consistent improvements over prior models in proactive response accuracy and efficiency, validating an effective solution for proactive video understanding under strict computational constraints.

cs.CV↗