arXiv subjects

Zhenhua Wu

Publications and source records attributed to Zhenhua Wu.

At least 19 recordsLinked to original sources

arXiv2026

RealWeather: Realistic and Scene-Faithful Weather Translation with Driving World Models

Realistic weather translation is valuable for developing and evaluating autonomous driving systems, yet collecting paired videos of the same scenes under different weather conditions at scale is impractical. Existing methods therefore rely on synthetic data, 3D weather editing, or geometry-conditioned generation, often compromising weather realism or scene fidelity. We propose RealWeather, a driving world model for both realistic and scene-faithful weather translation. Our key idea is to learn authentic weather dynamics directly from real-world videos. Specifically, RealWeather employs Progressive Realism Bootstrapping, an iterative data-refinement strategy. Assisted by an auxiliary Pseudo-Clear Generation pipeline, training initially starts with pseudo-style conditioning videos. As training proceeds, these inputs are progressively replaced with increasingly realistic videos generated by the model itself. This strategy bridges the pseudo-to-real domain gap, allowing the model to adapt seamlessly to real-world input distributions and naturally support bidirectional clear adverse translation. Furthermore, to strictly enforce structural integrity and suppress hallucinations, we introduce Scene-Fidelity RL Optimization, a reward-driven policy optimization strategy that explicitly penalizes alterations to safety-critical driving elements. Extensive experiments demonstrate that RealWeather significantly outperforms existing methods in visual realism and structural preservation, while enabling robust long-tail weather scenario generation and strong zero-shot out-of-distribution generalization. Our video demos can be found at https://hust-umi.github.io/RealWeather/.

cs.CV↗

arXiv2026

RealityBridge: Bridging Editable 3D Gaussian Splatting Driving Simulations and Real-World Videos

Long-tail hazardous scenarios are essential for safety-oriented autonomous driving, yet they are difficult to collect at scale. Editable 3D Gaussian Splatting (3DGS) simulation offers a scalable alternative through real-scene reconstruction and controllable editing. However, edited 3DGS-rendered videos often exhibit a significant Sim-to-Real gap, manifested as rendering artifacts, degraded foreground assets, illumination mismatch, and temporal flickering. Addressing these coupled defects requires jointly restoring local appearance, harmonizing edited content, and maintaining temporal consistency, whereas existing methods typically address only a subset of these requirements. To fill this gap, we propose RealityBridge, a video restoration and harmonization framework that converts edited 3DGS renderings into realistic driving footage while preserving simulator-defined structure, edits, and dynamics. RealityBridge conditions a video foundation model on complementary modality signals, with a lightweight GateNet adaptively controlling their injection across backbone blocks. We further develop a task-oriented curation pipeline to construct training data, and design a four-stage supervised training strategy followed by reward-guided post-training. Extensive experiments demonstrate that RealityBridge outperforms existing methods in restoration and harmonization while preserving strong temporal consistency.

cs.CV↗

arXiv2026

CacheClip: Accelerating RAG with Effective KV Cache Reuse

Retrieval-Augmented Generation (RAG) systems suffer from severe time-to-first-token (TTFT) bottlenecks due to long input sequences. Existing KV cache reuse methods face a fundamental trade-off: prefix caching requires identical prefixes that rarely occur in RAG scenarios, while direct precomputation sacrifices quality due to missing inter-chunk attention and repeated attention sinks. Recent methods like APE and CacheBlend partially address these issues but remain inadequate for robust RAG applications. This paper presents CacheClip, a novel framework that achieves both fast TTFT and high generation quality. Our key insight is that small auxiliary LLMs exhibit similar last-layer attention distributions to primary LLMs (the target model for generation), enabling efficient identification of tokens critical for restoring inter-chunk attention, thereby significantly improving response quality on cross-chunk reasoning tasks. CacheClip integrates four techniques: (1) auxiliary-model-guided token selection for selective KV cache recomputation, (2) shared prefixes to eliminate redundant attention sinks, (3) a sliding-window grouping strategy to maintain local coherence during partial KV cache updates, and (4) a CPU-GPU hybrid design that offloads auxiliary model inference to idle CPU resources, avoiding additional GPU overhead. The recomputation ratio is adjustable, allowing users to flexibly balance efficiency and quality for different deployment requirements. Experiments show CacheClip retains up to 85.2% and 91.1% of full-attention performance on NIAH and LongBench, outperforming CacheBlend and APE by 16.1 and 12.8 points on NIAH, and by 4.5 and 4.2 points on LongBench (with recomp% = 20%). Meanwhile, CacheClip accelerates LLM inference by up to 3.33$\times$ in prefill time (with recomp% = 20%), providing a practical solution to the efficiency-quality trade-off in RAG systems.

cs.LG↗

arXiv2026

LatentOmni: Rethinking Omni-Modal Understanding via Unified Audio-Visual Latent Reasoning

Joint audio-visual reasoning is essential for omnimodal understanding, yet current multimodal large language models (MLLMs) still struggle when reasoning requires fine-grained evidence from both modalities. A central limitation is that explicit text-based chain-of-thought (CoT) compresses continuous audio-visual signals into discrete tokens, weakening temporal grounding and shifting intermediate reasoning toward language priors. We argue that a unified latent space is a better medium for such reasoning because it preserves dense sensory information while remaining compatible with autoregressive generation. Based on this insight, we propose \textbf{LatentOmni}, a cross-modal reasoning framework that interleaves textual reasoning with audio-visual latent states. LatentOmni introduces feature-level supervision to align latent reasoning states with task-relevant sensory features and uses Omni-Sync Position Embedding (OSPE) to maintain temporal consistency between latent audio and visual states. We further construct \textbf{LatentOmni-Instruct-35K}, a dataset of audio-visual interleaved reasoning trajectories for supervising latent-space reasoning. Comprehensive evaluation across multiple audio-visual reasoning benchmarks demonstrates that LatentOmni achieves the best performance among the evaluated open-source models and consistently outperforms the Explicit Text CoT baseline, supporting latent-space joint reasoning as a promising path toward stronger omnimodal understanding.

cs.CL↗

arXiv2026

A ferroelectric junction transistor memory made from switchable van der Waals p-n heterojunctions

Van der Waals (vdW) p-n heterojunctions are important building blocks for advanced electronics and optoelectronics, in which high-quality heterojunctions essentially determine device performances or functionalities. Creating tunable depletion regions with substantially suppressed leakage currents presents huge challenges, but is crucial for heterojunction applications. Here, by using band-aligned p-type SnSe and n-type ferroelectric α-In2Se3 as a model, we report near-ideal multifunctional vdW p-n heterojunctions with small reverse leakage currents (0.1 pA) and a desired diode ideality factor (1.95). We realize ferroelectric-tuned band alignment with a giant barrier modulation of 900 meV. Based on such tunable heterojunctions, we propose and demonstrate a fundamental different memory device termed ferroelectric junction field-effect transistor memory, which shows large memory windows (1.8 V), ultrafast speed (100 ns), high operation temperature (393 K), and low cycle-to-cycle variation (2%). Additionally, the reliable synaptic characteristics of these memory devices promise low-power neuromorphic computing. Our work provides a new device platform with switchable memory heterojunctions, applicable to high performance brain-inspired electronics and optoelectronics.

cond-mat.mtrl-sci↗

arXiv2026

A Generalizable Framework for Building Executable Domain-Specific LLMs under Data Scarcity: Demonstration on Semiconductor TCAD Simulation

Scientific and engineering verticals often suffer from data scarcity and strict executability requirements: models must generate not only fluent text, but also syntactically valid, tool-compilable scripts. We present a schema-first alignment framework for building compact, executable domain-specific LLMs in low-resource settings. The framework integrates three core components: (i) large-scale synthetic QA data generation from expert documentation to instill foundational domain knowledge; (ii) a code-centric IR->DPO workflow that converts verified tool decks into interpretable intermediate representations (IR), performs equivalence-preserving diversification, and constructs preference pairs to directly optimize instruction compliance and code executability; and (iii) a controlled evaluation of Retrieval-Augmented Generation (RAG), showing that while RAG benefits general LLMs, it can marginally degrade the performance of already domain-aligned models. We demonstrate the framework by instantiating TcadGPT for semiconductor Technology Computer-Aided Design (TCAD). Using 1.5M synthetic QA pairs and an IR-driven DPO dataset, TcadGPT attains 85.6% semantic accuracy and an 80.0% syntax pass rate on SDE executability tests, substantially outperforming state-of-the-art general LLMs such as GPT-4o. To probe portability beyond TCAD, we apply the same recipe to the open-source FEM solver Elmer, observing consistent improvements in script-level success rates over general-purpose baselines. All datasets, benchmarks, and code (including P1, P2, and IR->DPO) are released for reproducibility. Together, these results suggest that the proposed framework provides a robust and reproducible path toward executable LLMs in specialized, data-scarce professional domains.

cs.CE↗

arXiv2025

Kling-Omni Technical Report

We present Kling-Omni, a generalist generative framework designed to synthesize high-fidelity videos directly from multimodal visual language inputs. Adopting an end-to-end perspective, Kling-Omni bridges the functional separation among diverse video generation, editing, and intelligent reasoning tasks, integrating them into a holistic system. Unlike disjointed pipeline approaches, Kling-Omni supports a diverse range of user inputs, including text instructions, reference images, and video contexts, processing them into a unified multimodal representation to deliver cinematic-quality and highly-intelligent video content creation. To support these capabilities, we constructed a comprehensive data system that serves as the foundation for multimodal video creation. The framework is further empowered by efficient large-scale pre-training strategies and infrastructure optimizations for inference. Comprehensive evaluations reveal that Kling-Omni demonstrates exceptional capabilities in in-context generation, reasoning-based editing, and multimodal instruction following. Moving beyond a content creation tool, we believe Kling-Omni is a pivotal advancement toward multimodal world simulators capable of perceiving, reasoning, generating and interacting with the dynamic and complex worlds.

cs.CV↗

arXiv2025

Mavors: Multi-granularity Video Representation for Multimodal Large Language Model

Long-context video understanding in multimodal large language models (MLLMs) faces a critical challenge: balancing computational efficiency with the retention of fine-grained spatio-temporal patterns. Existing approaches (e.g., sparse sampling, dense sampling with low resolution, and token compression) suffer from significant information loss in temporal dynamics, spatial details, or subtle interactions, particularly in videos with complex motion or varying resolutions. To address this, we propose $\mathbf{Mavors}$, a novel framework that introduces $\mathbf{M}$ulti-gr$\mathbf{a}$nularity $\mathbf{v}$ide$\mathbf{o}$ $\mathbf{r}$epre$\mathbf{s}$entation for holistic long-video modeling. Specifically, Mavors directly encodes raw video content into latent representations through two core components: 1) an Intra-chunk Vision Encoder (IVE) that preserves high-resolution spatial features via 3D convolutions and Vision Transformers, and 2) an Inter-chunk Feature Aggregator (IFA) that establishes temporal coherence across chunks using transformer-based dependency modeling with chunk-level rotary position encodings. Moreover, the framework unifies image and video understanding by treating images as single-frame videos via sub-image decomposition. Experiments across diverse benchmarks demonstrate Mavors' superiority in maintaining both spatial fidelity and temporal continuity, significantly outperforming existing methods in tasks requiring fine-grained spatio-temporal reasoning.

cs.CV↗

arXiv2025

Kwai Keye-VL 1.5 Technical Report

In recent years, the development of Large Language Models (LLMs) has significantly advanced, extending their capabilities to multimodal tasks through Multimodal Large Language Models (MLLMs). However, video understanding remains a challenging area due to the dynamic and information-dense nature of videos. Existing models struggle with the trade-off between spatial resolution and temporal coverage when processing video content. We present Keye-VL-1.5, which addresses fundamental challenges in video comprehension through three key innovations. First, we introduce a novel Slow-Fast video encoding strategy that dynamically allocates computational resources based on inter-frame similarity, processing key frames with significant visual changes at higher resolution (Slow pathway) while handling relatively static frames with increased temporal coverage at lower resolution (Fast pathway). Second, we implement a progressive four-stage pre-training methodology that systematically extends the model's context length from 8K to 128K tokens, enabling processing of longer videos and more complex visual content. Third, we develop a comprehensive post-training pipeline focusing on reasoning enhancement and human preference alignment, incorporating a 5-step chain-of-thought data construction process, iterative GSPO-based reinforcement learning with progressive prompt hinting for difficult cases, and alignment training. Through extensive evaluation on public benchmarks and rigorous internal human assessment, Keye-VL-1.5 demonstrates significant improvements over existing models, particularly excelling in video understanding tasks while maintaining competitive performance on general multimodal benchmarks.

cs.CV↗

arXiv2025

Organ-Agents: Virtual Human Physiology Simulator via LLMs

Recent advances in large language models (LLMs) have enabled new possibilities in simulating complex physiological systems. We introduce Organ-Agents, a multi-agent framework that simulates human physiology via LLM-driven agents. Each Simulator models a specific system (e.g., cardiovascular, renal, immune). Training consists of supervised fine-tuning on system-specific time-series data, followed by reinforcement-guided coordination using dynamic reference selection and error correction. We curated data from 7,134 sepsis patients and 7,895 controls, generating high-resolution trajectories across 9 systems and 125 variables. Organ-Agents achieved high simulation accuracy on 4,509 held-out patients, with per-system MSEs <0.16 and robustness across SOFA-based severity strata. External validation on 22,689 ICU patients from two hospitals showed moderate degradation under distribution shifts with stable simulation. Organ-Agents faithfully reproduces critical multi-system events (e.g., hypotension, hyperlactatemia, hypoxemia) with coherent timing and phase progression. Evaluation by 15 critical care physicians confirmed realism and physiological plausibility (mean Likert ratings 3.9 and 3.7). Organ-Agents also enables counterfactual simulations under alternative sepsis treatment strategies, generating trajectories and APACHE II scores aligned with matched real-world patients. In downstream early warning tasks, classifiers trained on synthetic data showed minimal AUROC drops (<0.04), indicating preserved decision-relevant patterns. These results position Organ-Agents as a credible, interpretable, and generalizable digital twin for precision diagnosis, treatment simulation, and hypothesis testing in critical care.

cs.LG↗

arXiv2025

Kwai Keye-VL Technical Report

While Multimodal Large Language Models (MLLMs) demonstrate remarkable capabilities on static images, they often fall short in comprehending dynamic, information-dense short-form videos, a dominant medium in today's digital landscape. To bridge this gap, we introduce \textbf{Kwai Keye-VL}, an 8-billion-parameter multimodal foundation model engineered for leading-edge performance in short-video understanding while maintaining robust general-purpose vision-language abilities. The development of Keye-VL rests on two core pillars: a massive, high-quality dataset exceeding 600 billion tokens with a strong emphasis on video, and an innovative training recipe. This recipe features a four-stage pre-training process for solid vision-language alignment, followed by a meticulous two-phase post-training process. The first post-training stage enhances foundational capabilities like instruction following, while the second phase focuses on stimulating advanced reasoning. In this second phase, a key innovation is our five-mode ``cold-start'' data mixture, which includes ``thinking'', ``non-thinking'', ``auto-think'', ``think with image'', and high-quality video data. This mixture teaches the model to decide when and how to reason. Subsequent reinforcement learning (RL) and alignment steps further enhance these reasoning capabilities and correct abnormal model behaviors, such as repetitive outputs. To validate our approach, we conduct extensive evaluations, showing that Keye-VL achieves state-of-the-art results on public video benchmarks and remains highly competitive on general image-based tasks (Figure 1). Furthermore, we develop and release the \textbf{KC-MMBench}, a new benchmark tailored for real-world short-video scenarios, where Keye-VL shows a significant advantage.

cs.CV↗

arXiv2025

Low-Rank Adaptation of Pre-Trained Stable Diffusion for Rigid-Body Target ISAR Imaging

Traditional range-instantaneous Doppler (RID) methods for rigid-body target imaging often suffer from low resolution due to the limitations of time-frequency analysis (TFA). To address this challenge, our primary focus is on obtaining high resolution time-frequency representations (TFRs) from their low resolution counterparts. Recognizing that the curve features of TFRs are a specific type of texture feature, we argue that pre trained generative models such as Stable Diffusion (SD) are well suited for enhancing TFRs, thanks to their powerful capability in capturing texture representations. Building on this insight, we propose a novel inverse synthetic aperture radar (ISAR) imaging method for rigid-body targets, leveraging the low-rank adaptation (LoRA) of a pre-trained SD model. Our approach adopts the basic structure and pre-trained parameters of SD Turbo while incorporating additional linear operations for LoRA and adversarial training to achieve super-resolution and noise suppression. Then we integrate LoRA-SD into the RID-based ISAR imaging, enabling sharply focused and denoised imaging with super-resolution capabilities. We evaluate our method using both simulated and real radar data. The experimental results demonstrate the superiority of our approach in frequency es timation and ISAR imaging compared to traditional methods. Notably, the generalization capability is verified by training on simulated radar data and testing on measured radar data.

cs.CV↗

arXiv2025

MTGS: Multi-Traversal Gaussian Splatting

Multi-traversal data, commonly collected through daily commutes or by self-driving fleets, provides multiple viewpoints for scene reconstruction within a road block. This data offers significant potential for high-quality novel view synthesis, which is crucial for applications such as autonomous vehicle simulators. However, inherent challenges in multi-traversal data often result in suboptimal reconstruction quality, including variations in appearance and the presence of dynamic objects. To address these issues, we propose Multi-Traversal Gaussian Splatting (MTGS), a novel approach that reconstructs high-quality driving scenes from arbitrarily collected multi-traversal data by modeling a shared static geometry while separately handling dynamic elements and appearance variations. Our method employs a multi-traversal dynamic scene graph with a shared static node and traversal-specific dynamic nodes, complemented by color correction nodes with learnable spherical harmonics coefficient residuals. This approach enables high-fidelity novel view synthesis and provides flexibility to navigate any viewpoint. We conduct extensive experiments on a large-scale driving dataset, nuPlan, with multi-traversal data. Our results demonstrate that MTGS improves LPIPS by 23.5% and geometry accuracy by 46.3% compared to single-traversal baselines. The code and data would be available to the public.

cs.CV↗

arXiv2024

Gravitational Wave-Sensitive Photonic-Like Electronic Transport in Graphene for Efficient High-Frequency Gravitational Wave Detection

High-frequency gravitational waves are crucial for understanding the very early universe and distinguishing between various cosmological models, but detecting them remains a significant challenge. We investigated the effects of high-frequency gravitational waves on photonic-like electronic transport in graphene. The results show that, unlike the influence of gravitational waves on the propagation of light, the influence of gravitational waves on photonic-like electronic transport can accumulate not only in real space but also in $k$-space. This makes photonic-like electronic transport under gravitational waves similar to the propagation of light in a medium where the refractive index varies dramatically due to gravitational waves, and with shorter wavelengths. As a result, the relative intensity variation in photonic-like electronic transport under gravitational waves exceeds that of a laser interferometer with the same arm length by six orders of magnitude. At low temperatures, the influence of phonons on photon-like transport in the context of high-frequency gravitational waves can be ignored. These findings indicate a strong interaction between gravitational waves and electron transport, which helps to deepen the understanding of the interaction between gravitational waves and matter, and provides a different method for detecting high-frequency gravitational waves.

physics.ins-det↗

arXiv2024

Remote Sensing Image Segmentation Using Vision Mamba and Multi-Scale Multi-Frequency Feature Fusion

As remote sensing imaging technology continues to advance and evolve, processing high-resolution and diversified satellite imagery to improve segmentation accuracy and enhance interpretation efficiency emerg as a pivotal area of investigation within the realm of remote sensing. Although segmentation algorithms based on CNNs and Transformers achieve significant progress in performance, balancing segmentation accuracy and computational complexity remains challenging, limiting their wide application in practical tasks. To address this, this paper introduces state space model (SSM) and proposes a novel hybrid semantic segmentation network based on vision Mamba (CVMH-UNet). This method designs a cross-scanning visual state space block (CVSSBlock) that uses cross 2D scanning (CS2D) to fully capture global information from multiple directions, while by incorporating convolutional neural network branches to overcome the constraints of Vision Mamba (VMamba) in acquiring local information, this approach facilitates a comprehensive analysis of both global and local features. Furthermore, to address the issue of limited discriminative power and the difficulty in achieving detailed fusion with direct skip connections, a multi-frequency multi-scale feature fusion block (MFMSBlock) is designed. This module introduces multi-frequency information through 2D discrete cosine transform (2D DCT) to enhance information utilization and provides additional scale local detail information through point-wise convolution branches. Finally, it aggregates multi-scale information along the channel dimension, achieving refined feature fusion. Findings from experiments conducted on renowned datasets of remote sensing imagery demonstrate that proposed CVMH-UNet achieves superior segmentation performance while maintaining low computational complexity, outperforming surpassing current leading-edge segmentation algorithms.

cs.CV↗

arXiv2024

ToDER: Towards Colonoscopy Depth Estimation and Reconstruction with Geometry Constraint Adaptation

Visualizing colonoscopy is crucial for medical auxiliary diagnosis to prevent undetected polyps in areas that are not fully observed. Traditional feature-based and depth-based reconstruction approaches usually end up with undesirable results due to incorrect point matching or imprecise depth estimation in realistic colonoscopy videos. Modern deep-based methods often require a sufficient number of ground truth samples, which are generally hard to obtain in optical colonoscopy. To address this issue, self-supervised and domain adaptation methods have been explored. However, these methods neglect geometry constraints and exhibit lower accuracy in predicting detailed depth. We thus propose a novel reconstruction pipeline with a bi-directional adaptation architecture named ToDER to get precise depth estimations. Furthermore, we carefully design a TNet module in our adaptation architecture to yield geometry constraints and obtain better depth quality. Estimated depth is finally utilized to reconstruct a reliable colon model for visualization. Experimental results demonstrate that our approach can precisely predict depth maps in both realistic and synthetic colonoscopy videos compared with other self-supervised and domain adaptation methods. Our method on realistic colonoscopy also shows the great potential for visualizing unobserved regions and preventing misdiagnoses.

cs.CV↗

arXiv2024

Nanofabrication beyond optical diffraction limit: Optical driven assembly enabled by superlubricity

The optical manipulation of nanoparticles on superlubricity surfaces is investigated. The research revealed that, due to the near-zero static friction and extremely low dynamic friction at superlubricity interfaces, the maximum intensity for controlling the optical field can be less than 100 W/cm$^2$, which is nine orders of magnitude lower than controlling nanoparticles on traditional interfaces. The controlled nanoparticle radius can be as small as 5 nm, which is more than one order of magnitude smaller than nanoparticles controlled through traditional optical manipulation. Manipulation can be achieved in sub-microsecond to microsecond timescales. Furthermore, the manipulation takes place on solid surfaces and in non-liquid environments, with minimal impact from Brownian motion. By appropriately increasing dynamic friction, controlling light intensity, or reducing pressure, the effects of Brownian motion can be eliminated, allowing for the construction of microstructures with a size as small as 1/75 of the wavelength of light. This enables the control of super-resolution optical microstructures. The optical super-resolution manipulation of nanoparticles on superlubricity surfaces will find important applications in fields such as nanofabrication, photolithography, optical metasurface, and biochemical analysis.

physics.app-ph↗

arXiv2023

High-sensitivity graphene MEMS force and acceleration sensor based on graphene-induced non-radiative transition

The micro-electromechanical-system (MEMS) force and acceleration sensor utilizing the graphene-induced non-radiative transition was investigated. The graphene-induced non-radiative transition is very sensitive to the distance, and the deflection of the graphene ribbon is highly susceptive to applied force or acceleration. Thus, a high-sensitivity MEMS sensor can be achieved with detecting the graphene ribbon's deflection of 1 nm, the force of 0.1 pN, and the acceleration of 0.1 mg. The MEMS sensor, with a size of only tens of microns, can be charged by light irradiation without connecting power sources. In addition, it allows long-distance detection, i.e., wireless transmitter circuit can be omitted. Therefore, it will have significant application prospects in the fields of micro-smart devices, wearable devices, biomedical systems, and so on.

cond-mat.mes-hall↗