arXiv subjects

Kai Wang

Publications and source records attributed to Kai Wang.

At least 361 records · Page 20Linked to original sources

arXiv2025

InterLCM: Low-Quality Images as Intermediate States of Latent Consistency Models for Effective Blind Face Restoration

Diffusion priors have been used for blind face restoration (BFR) by fine-tuning diffusion models (DMs) on restoration datasets to recover low-quality images. However, the naive application of DMs presents several key limitations. (i) The diffusion prior has inferior semantic consistency (e.g., ID, structure and color.), increasing the difficulty of optimizing the BFR model; (ii) reliance on hundreds of denoising iterations, preventing the effective cooperation with perceptual losses, which is crucial for faithful restoration. Observing that the latent consistency model (LCM) learns consistency noise-to-data mappings on the ODE-trajectory and therefore shows more semantic consistency in the subject identity, structural information and color preservation, we propose InterLCM to leverage the LCM for its superior semantic consistency and efficiency to counter the above issues. Treating low-quality images as the intermediate state of LCM, InterLCM achieves a balance between fidelity and quality by starting from earlier LCM steps. LCM also allows the integration of perceptual loss during training, leading to improved restoration quality, particularly in real-world scenarios. To mitigate structural and semantic uncertainties, InterLCM incorporates a Visual Module to extract visual features and a Spatial Encoder to capture spatial details, enhancing the fidelity of restored images. Extensive experiments demonstrate that InterLCM outperforms existing approaches in both synthetic and real-world datasets while also achieving faster inference speed.

cs.CV↗

arXiv2025

ATHENA: An In-vehicle CAN Intrusion Detection Framework Based on Physical Characteristics of Vehicle Systems

With the growing interconnection between In-Vehicle Networks (IVNs) and external environments, intelligent vehicles are increasingly vulnerable to sophisticated external network attacks. This paper proposes ATHENA, the first IVN intrusion detection framework that adopts a vehicle-cloud integrated architecture to achieve better security performance for the resource-constrained vehicular environment. Specifically, in the cloud with sufficient resources, ATHENA uses the clustering method of multi-distribution mixture model combined with deep data mining technology to generate the raw Payload Rule Bank of IVN CAN messages, and then improves the rule quality with the help of exploitation on the first-principled physical knowledge of the vehicle system, after which the payload rules are periodically sent to the vehicle terminal. At the vehicle terminal, a simple LSTM component is used to generate the Time Rule Bank representing the long-term time series dependencies and the periodic characteristics of CAN messages, but not for any detection tasks as in traditional usage scenarios, where only the generated time rules are the candidates for further IVN intrusion detection tasks. Based on both the payload and time rules generated from cloud and vehicle terminal, ATHENA can achieve efficient intrusion detection capability by simple rule-base matching operations, rather than using complex black-box reasoning of resource-intensive neural network models, which is in fact only used for rule logic generation phase instead of the actual intrusion detection phase in our framework. Comparative experimental results on the ROAD dataset, which is current the most outstanding real-world in-vehicle CAN dataset covering new instances of sophisticated and stealthy masquerade attacks, demonstrate ATHENA significantly outperforms the state-of-the-art IVN intrusion detection methods in detecting complex attacks.

cs.CR↗

arXiv2025

Causal Deciphering and Inpainting in Spatio-Temporal Dynamics via Diffusion Model

Spatio-temporal (ST) prediction has garnered a De facto attention in earth sciences, such as meteorological prediction, human mobility perception. However, the scarcity of data coupled with the high expenses involved in sensor deployment results in notable data imbalances. Furthermore, models that are excessively customized and devoid of causal connections further undermine the generalizability and interpretability. To this end, we establish a causal framework for ST predictions, termed CaPaint, which targets to identify causal regions in data and endow model with causal reasoning ability in a two-stage process. Going beyond this process, we utilize the back-door adjustment to specifically address the sub-regions identified as non-causal in the upstream phase. Specifically, we employ a novel image inpainting technique. By using a fine-tuned unconditional Diffusion Probabilistic Model (DDPM) as the generative prior, we in-fill the masks defined as environmental parts, offering the possibility of reliable extrapolation for potential data distributions. CaPaint overcomes the high complexity dilemma of optimal ST causal discovery models by reducing the data generation complexity from exponential to quasi-linear levels. Extensive experiments conducted on five real-world ST benchmarks demonstrate that integrating the CaPaint concept allows models to achieve improvements ranging from 4.3% to 77.3%. Moreover, compared to traditional mainstream ST augmenters, CaPaint underscores the potential of diffusion models in ST enhancement, offering a novel paradigm for this field. Our project is available at https://anonymous.4open.science/r/12345-DFCC.

cs.CV↗

arXiv2025

Distilling Long-tailed Datasets

Dataset distillation aims to synthesize a small, information-rich dataset from a large one for efficient model training. However, existing dataset distillation methods struggle with long-tailed datasets, which are prevalent in real-world scenarios. By investigating the reasons behind this unexpected result, we identified two main causes: 1) The distillation process on imbalanced datasets develops biased gradients, leading to the synthesis of similarly imbalanced distilled datasets. 2) The experts trained on such datasets perform suboptimally on tail classes, resulting in misguided distillation supervision and poor-quality soft-label initialization. To address these issues, we first propose Distribution-agnostic Matching to avoid directly matching the biased expert trajectories. It reduces the distance between the student and the biased expert trajectories and prevents the tail class bias from being distilled to the synthetic dataset. Moreover, we improve the distillation guidance with Expert Decoupling, which jointly matches the decoupled backbone and classifier to improve the tail class performance and initialize reliable soft labels. This work pioneers the field of long-tailed dataset distillation, marking the first effective effort to distill long-tailed datasets.

cs.LG↗

arXiv2025

Ten-channel Hong-Ou-Mandel interference between independent optical combs

Dissipative Kerr soliton (DKS) frequency comb exhibits broad and narrow-linewidth frequency modes, which make it suitable for quantum communication. However, scalable quantum network based on multiple independent combs is still a challenge due to their fabrication-induced frequency mismatches. This limitation becomes critical in measurement-device-independent quantum key distribution, which requires high visibility of Hong-Ou-Mandel interference between multiple frequency channels. Here, we experimentally demonstrate two independent DKS combs with ten spectrally aligned lines without any frequency locking system. The visibility for individual comb-line pairs reaches up to $46.72 \pm 0.63\%$ via precision frequency translation, establishing a foundation for deploying DKS combs in multi-user quantum networks.

quant-ph↗

arXiv2025

MPBench: A Comprehensive Multimodal Reasoning Benchmark for Process Errors Identification

Reasoning is an essential capacity for large language models (LLMs) to address complex tasks, where the identification of process errors is vital for improving this ability. Recently, process-level reward models (PRMs) were proposed to provide step-wise rewards that facilitate reinforcement learning and data production during training and guide LLMs toward correct steps during inference, thereby improving reasoning accuracy. However, existing benchmarks of PRMs are text-based and focus on error detection, neglecting other scenarios like reasoning search. To address this gap, we introduce MPBench, a comprehensive, multi-task, multimodal benchmark designed to systematically assess the effectiveness of PRMs in diverse scenarios. MPBench employs three evaluation paradigms, each targeting a specific role of PRMs in the reasoning process: (1) Step Correctness, which assesses the correctness of each intermediate reasoning step; (2) Answer Aggregation, which aggregates multiple solutions and selects the best one; and (3) Reasoning Process Search, which guides the search for optimal reasoning steps during inference. Through these paradigms, MPBench makes comprehensive evaluations and provides insights into the development of multimodal PRMs.

cs.AI↗

arXiv2025

Make Optimization Once and for All with Fine-grained Guidance

Learning to Optimize (L2O) enhances optimization efficiency with integrated neural networks. L2O paradigms achieve great outcomes, e.g., refitting optimizer, generating unseen solutions iteratively or directly. However, conventional L2O methods require intricate design and rely on specific optimization processes, limiting scalability and generalization. Our analyses explore general framework for learning optimization, called Diff-L2O, focusing on augmenting sampled solutions from a wider view rather than local updates in real optimization process only. Meanwhile, we give the related generalization bound, showing that the sample diversity of Diff-L2O brings better performance. This bound can be simply applied to other fields, discussing diversity, mean-variance, and different tasks. Diff-L2O's strong compatibility is empirically verified with only minute-level training, comparing with other hour-levels.

cs.LG↗

arXiv2025

Optimizing for the Shortest Path in Denoising Diffusion Model

In this research, we propose a novel denoising diffusion model based on shortest-path modeling that optimizes residual propagation to enhance both denoising efficiency and quality. Drawing on Denoising Diffusion Implicit Models (DDIM) and insights from graph theory, our model, termed the Shortest Path Diffusion Model (ShortDF), treats the denoising process as a shortest-path problem aimed at minimizing reconstruction error. By optimizing the initial residuals, we improve the efficiency of the reverse diffusion process and the quality of the generated samples. Extensive experiments on multiple standard benchmarks demonstrate that ShortDF significantly reduces diffusion time (or steps) while enhancing the visual fidelity of generated samples compared to prior arts. This work, we suppose, paves the way for interactive diffusion-based applications and establishes a foundation for rapid data generation. Code is available at https://github.com/UnicomAI/ShortDF.

cs.CV↗

arXiv2025

An Improved Two-Step Attack on Lattice-Based Cryptography: A Case Study of Kyber

After three rounds of post-quantum cryptography (PQC) strict evaluations conducted by NIST, CRYSTALS-Kyber was successfully selected in July 2022 and standardized in August 2024. It becomes urgent to further evaluate Kyber's physical security for the upcoming deployment phase. In this brief, we present an improved two-step attack on Kyber to quickly recover the full secret key, s, by using much fewer power traces and less time. In the first step, we use the correlation power analysis (CPA) to obtain a portion of guess values of s with a small number of power traces. The CPA is enhanced by utilizing both Pearson and Kendall's rank correlation coefficients and modifying the leakage model to improve the accuracy. In the second step, we adopt the lattice attack to recover s based on the results of CPA. The success rate is largely built up by constructing a trial-and-error method. We deploy the reference implementations of Kyber-512, -768, and -1024 on an ARM Cortex-M4 target board and successfully recover s in approximately 9~10 minutes with at most 15 power traces, using a Xeon Gold 6342-equipped machine for the attack.

cs.CR↗

arXiv2025

HunyuanVideo: A Systematic Framework For Large Video Generative Models

Recent advancements in video generation have significantly impacted daily life for both individuals and industries. However, the leading video generation models remain closed-source, resulting in a notable performance gap between industry capabilities and those available to the public. In this report, we introduce HunyuanVideo, an innovative open-source video foundation model that demonstrates performance in video generation comparable to, or even surpassing, that of leading closed-source models. HunyuanVideo encompasses a comprehensive framework that integrates several key elements, including data curation, advanced architectural design, progressive model scaling and training, and an efficient infrastructure tailored for large-scale model training and inference. As a result, we successfully trained a video generative model with over 13 billion parameters, making it the largest among all open-source models. We conducted extensive experiments and implemented a series of targeted designs to ensure high visual quality, motion dynamics, text-video alignment, and advanced filming techniques. According to evaluations by professionals, HunyuanVideo outperforms previous state-of-the-art models, including Runway Gen-3, Luma 1.6, and three top-performing Chinese video generative models. By releasing the code for the foundation model and its applications, we aim to bridge the gap between closed-source and open-source communities. This initiative will empower individuals within the community to experiment with their ideas, fostering a more dynamic and vibrant video generation ecosystem. The code is publicly available at https://github.com/Tencent/HunyuanVideo.

cs.CV↗

arXiv2025

X-Field: A Physically Grounded Representation for 3D X-ray Reconstruction

X-ray imaging is indispensable in medical diagnostics, yet its use is tightly regulated due to potential health risks. To mitigate radiation exposure, recent research focuses on generating novel views from sparse inputs and reconstructing Computed Tomography (CT) volumes, borrowing representations from the 3D reconstruction area. However, these representations originally target visible light imaging that emphasizes reflection and scattering effects, while neglecting penetration and attenuation properties of X-ray imaging. In this paper, we introduce X-Field, the first 3D representation specifically designed for X-ray imaging, rooted in the energy absorption rates across different materials. To accurately model diverse materials within internal structures, we employ 3D ellipsoids with distinct attenuation coefficients. To estimate each material's energy absorption of X-rays, we devise an efficient path partitioning algorithm accounting for complex ellipsoid intersections. We further propose hybrid progressive initialization to refine the geometric accuracy of X-Filed and incorporate material-based optimization to enhance model fitting along material boundaries. Experiments show that X-Field achieves superior visual fidelity on both real-world human organ and synthetic object datasets, outperforming state-of-the-art methods in X-ray Novel View Synthesis and CT Reconstruction.

cs.CV↗

arXiv2025

A Multimodal Benchmark Dataset and Model for Crop Disease Diagnosis

While conversational generative AI has shown considerable potential in enhancing decision-making for agricultural professionals, its exploration has predominantly been anchored in text-based interactions. The evolution of multimodal conversational AI, leveraging vast amounts of image-text data from diverse sources, marks a significant stride forward. However, the application of such advanced vision-language models in the agricultural domain, particularly for crop disease diagnosis, remains underexplored. In this work, we present the crop disease domain multimodal (CDDM) dataset, a pioneering resource designed to advance the field of agricultural research through the application of multimodal learning techniques. The dataset comprises 137,000 images of various crop diseases, accompanied by 1 million question-answer pairs that span a broad spectrum of agricultural knowledge, from disease identification to management practices. By integrating visual and textual data, CDDM facilitates the development of sophisticated question-answering systems capable of providing precise, useful advice to farmers and agricultural professionals. We demonstrate the utility of the dataset by finetuning state-of-the-art multimodal models, showcasing significant improvements in crop disease diagnosis. Specifically, we employed a novel finetuning strategy that utilizes low-rank adaptation (LoRA) to finetune the visual encoder, adapter and language model simultaneously. Our contributions include not only the dataset but also a finetuning strategy and a benchmark to stimulate further research in agricultural technology, aiming to bridge the gap between advanced AI techniques and practical agricultural applications. The dataset is available at https: //github.com/UnicomAI/UnicomBenchmark/tree/main/CDDMBench.

cs.CV↗

arXiv2025

Symplectic tracking through curved three dimensional fields by a method of generating functions

Symplectic integrator plays a pivotal role in the long-term tracking of charged particles within accelerators. To get symplectic maps in accurate simulation of single-particle trajectories, two key components are addressed: precise analytical expressions for arbitrary electromagnetic fields and a robust treatment of the equations of motion. In a source-free region, the electromagnetic fields can be decomposed into harmonic functions, applicable to both scalar and vector potentials, encompassing both straight and curved reference trajectories. These harmonics are constructed according to the boundary surface's field data due to uniqueness theorem. Finding generating functions to meet the Hamilton-Jacobi equation via a perturbative ansatz, we derive symplectic maps that are independent of the expansion order. This method yields a direct mapping from initial to final coordinates in a large step, bypassing the transition of intermediate coordinates. Our developed particle-tracking algorithm translates the field harmonics into beam trajectories, offering a high-resolution integration method in accelerator simulations.

physics.acc-ph↗

arXiv2025

Electron spin dynamics guide cell motility

Diverse organisms exploit the geomagnetic field (GMF) for migration. Migrating birds employ an intrinsically quantum mechanical mechanism for detecting the geomagnetic field: absorption of a blue photon generates a radical pair whose two electrons precess at different rates in the magnetic field, thereby sensitizing cells to the direction of the GMF. In this work, using an in vitro injury model, we discovered a quantum-based mechanism of cellular migration. Specifically, we show that migrating cells detect the GMF via an optically activated, electron spin-based mechanism. Cell injury provokes acute emission of blue photons, and these photons sensitize muscle progenitor cells to the magnetic field. We show that the magnetosensitivity of muscle progenitor cells is (a) activated by blue light, but not by green or red light, and (b) disrupted by the application of an oscillatory field at the frequency corresponding to the energy of the electron-spin/magnetic field interaction. A comprehensive analysis of protein expression reveals that the ability of blue photons to promote cell motility is mediated by activation of calmodulin calcium sensors. Collectively, these data suggest that cells possess a light-dependent magnetic compass driven by electron spin dynamics.

physics.bio-ph↗

arXiv2025

Deformation theory and Koszul duality for Rota-Baxter systems

This paper investigates Rota-Baxter systems in the sense of Brzeziński from the perspective of operad theory. The minimal model of the Rota-Baxter system operad is constructed, equivalently a concrete construction of its Koszul dual homotopy cooperad is given. The concept of homotopy Rota-Baxter systems and the $L_\infty$-algebra that governs deformations of a Rota-Baxter system are derived from the Koszul dual homotopy cooperad. The notion of infinity-Yang-Baxter pairs is introduced, which is a higher-order generalization of the traditional Yang-Baxter pairs. It is shown that a homotopy Rota-Baxter system structure on the endomorphism algebra of a graded space is equivalent to an associative infinity-Yang-Baxter pair on this graded algebra, thereby generalizing the classical correspondence between Yang-Baxter pairs and Rota-Baxter systems.

math.RA↗

arXiv2025

Digital Player: Evaluating Large Language Models based Human-like Agent in Games

With the rapid advancement of Large Language Models (LLMs), LLM-based autonomous agents have shown the potential to function as digital employees, such as digital analysts, teachers, and programmers. In this paper, we develop an application-level testbed based on the open-source strategy game "Unciv", which has millions of active players, to enable researchers to build a "data flywheel" for studying human-like agents in the "digital players" task. This "Civilization"-like game features expansive decision-making spaces along with rich linguistic interactions such as diplomatic negotiations and acts of deception, posing significant challenges for LLM-based agents in terms of numerical reasoning and long-term planning. Another challenge for "digital players" is to generate human-like responses for social interaction, collaboration, and negotiation with human players. The open-source project can be found at https:/github.com/fuxiAIlab/CivAgent.

cs.LG↗

arXiv2025

Real-Time Video Generation with Pyramid Attention Broadcast

We present Pyramid Attention Broadcast (PAB), a real-time, high quality and training-free approach for DiT-based video generation. Our method is founded on the observation that attention difference in the diffusion process exhibits a U-shaped pattern, indicating significant redundancy. We mitigate this by broadcasting attention outputs to subsequent steps in a pyramid style. It applies different broadcast strategies to each attention based on their variance for best efficiency. We further introduce broadcast sequence parallel for more efficient distributed inference. PAB demonstrates up to 10.5x speedup across three models compared to baselines, achieving real-time generation for up to 720p videos. We anticipate that our simple yet effective method will serve as a robust baseline and facilitate future research and application for video generation.

cs.CV↗

arXiv2025

Enhance-A-Video: Better Generated Video for Free

DiT-based video generation has achieved remarkable results, but research into enhancing existing models remains relatively unexplored. In this work, we introduce a training-free approach to enhance the coherence and quality of DiT-based generated videos, named Enhance-A-Video. The core idea is enhancing the cross-frame correlations based on non-diagonal temporal attention distributions. Thanks to its simple design, our approach can be easily applied to most DiT-based video generation frameworks without any retraining or fine-tuning. Across various DiT-based video generation models, our approach demonstrates promising improvements in both temporal consistency and visual quality. We hope this research can inspire future explorations in video generation enhancement.

cs.CV↗