arXiv subjects

Ke Wang

Publications and source records attributed to Ke Wang.

At least 253 records · Page 14Linked to original sources

arXiv2025

EFFOcc: Learning Efficient Occupancy Networks from Minimal Labels for Autonomous Driving

3D occupancy prediction (3DOcc) is a rapidly rising and challenging perception task in the field of autonomous driving. Existing 3D occupancy networks (OccNets) are both computationally heavy and label-hungry. In terms of model complexity, OccNets are commonly composed of heavy Conv3D modules or transformers at the voxel level. Moreover, OccNets are supervised with expensive large-scale dense voxel labels. Model and data inefficiencies, caused by excessive network parameters and label annotation requirements, severely hinder the onboard deployment of OccNets. This paper proposes an EFFicient Occupancy learning framework, EFFOcc, that targets minimal network complexity and label requirements while achieving state-of-the-art accuracy. We first propose an efficient fusion-based OccNet that only uses simple 2D operators and improves accuracy to the state-of-the-art on three large-scale benchmarks: Occ3D-nuScenes, Occ3D-Waymo, and OpenOccupancy-nuScenes. On the Occ3D-nuScenes benchmark, the fusion-based model with ResNet-18 as the image backbone has 21.35M parameters and achieves 51.49 in terms of mean Intersection over Union (mIoU). Furthermore, we propose a multi-stage occupancy-oriented distillation to efficiently transfer knowledge to vision-only OccNet. Extensive experiments on occupancy benchmarks show state-of-the-art precision for both fusion-based and vision-based OccNets. For the demonstration of learning with limited labels, we achieve 94.38\% of the performance (mIoU = 28.38) of a 100\% labeled vision OccNet (mIoU = 30.07) using the same OccNet trained with only 40\% labeled sequences and distillation from the fusion-based OccNet.

cs.CV↗

arXiv2025

The Higgs-Amplitude mode in the optical conductivity in the presence of a supercurrent: Gauge-invariant formulation with disorder

Observing the ``Higgs" or amplitude mode in superconductors has been a central challenge in condensed matter physics. Moreover, arriving at a theoretical understanding of this mode and how it is accessible in, say, conductivity experiments presents an additional challenge as here one needs to satisfy gauge invariance in the presence of disorder. In this paper, we characterize the Higgs contribution within a fully gauge-invariant treatment of the linear optical conductivity, $σ(ω)$, for a disordered superconductor carrying a uniform supercurrent. As a consequence of gauge invariance, there are two distinct charge conservation laws underlying the linear electromagnetic response with two associated sets of $f$-sum rules. An interesting finding from the Higgs-related sum rule is that the imaginary part of $σ(ω)$ yields an anisotropic, \textit{negative} $1/ω$ contribution in the THz regime. This is relevant to device applications and appears to be consistent with recent experiments. The work presented here emphasizes how difficult it is to disentangle the neutral amplitude mode contributions from those of the charged quasi-particles and we demonstrate why this is the case.

cond-mat.supr-con↗

arXiv2025

BounTCHA: A CAPTCHA Utilizing Boundary Identification in Guided Generative AI-extended Videos

In recent years, the rapid development of artificial intelligence (AI) especially multi-modal Large Language Models (MLLMs), has enabled it to understand text, images, videos, and other multimedia data, allowing AI systems to execute various tasks based on human-provided prompts. However, AI-powered bots have increasingly been able to bypass most existing CAPTCHA systems, posing significant security threats to web applications. This makes the design of new CAPTCHA mechanisms an urgent priority. We observe that humans are highly sensitive to shifts and abrupt changes in videos, while current AI systems still struggle to comprehend and respond to such situations effectively. Based on this observation, we design and implement BounTCHA, a CAPTCHA mechanism that leverages human perception of boundaries in video transitions and disruptions. By utilizing generative AI's capability to extend original videos with prompts, we introduce unexpected twists and changes to create a pipeline for generating guided short videos for CAPTCHA purposes. We develop a prototype and conduct experiments to collect data on humans' time biases in boundary identification. This data serves as a basis for distinguishing between human users and bots. Additionally, we perform a detailed security analysis of BounTCHA, demonstrating its resilience against various types of attacks. We hope that BounTCHA will act as a robust defense, safeguarding millions of web applications in the AI-driven era.

cs.CR↗

arXiv2025

Nonlinear Stability of Large-Period Traveling Waves Bifurcating from the Heteroclinic Loop in the FitzHugh-Nagumo Equation

A wave front and a wave back that spontaneously connect two hyperbolic equilibria, known as a heteroclinic wave loop, give rise to periodic waves with arbitrarily large spatial periods through the heteroclinic bifurcation. The nonlinear stability of these periodic waves is established in the setting of the FitzHugh-Nagumo equation, which is a well-known reaction-diffusion model with degenerate diffusion. First, for general systems, we give the expressions of spectra with small modulus for linearized operators about these periodic waves via the Lyapunov-Schmidt reduction and the Lin-Sandstede method. Second, applying these spectral results to the FitzHugh-Nagumo equation, we establish their diffusive spectral stability. Finally, we consider the nonlinear stability of these periodic waves against localized perturbations. We introduce a spatiotemporal phase modulation $φ$, and couple it with the associated modulated perturbation $\mathbf{V}$ along with the unmodulated perturbation $\mathbf{\widetilde{V}}$ to close a nonlinear iteration argument.

math.AP↗

arXiv2025

A Recent Supermassive Black Hole Binary in the Galactic Center Unveiled by the Hypervelocity Stars

When a binary of early-type stars from the young stellar populations in the Galactic center (GC) region is scattered to the vicinity of the supermassive black hole (SMBH) Sgr~$\rm{A}^{*}$, one of the components would be tidally ejected as an early-type hypervelocity star (HVS) and the counterpart would be captured on a tight orbit around Sgr~$\rm{A}^{*}$. Dozens of B-type HVSs moving faster than the Galactic escape speed have been discovered in the Galactic halo and are produced most likely by the SMBH Sgr~$\rm{A}^{*}$. However, the velocity distribution and in particular the deficit of the HVSs above $700\, \rm{km\, s^{-1}}$ is seriously inconsistent with the expectations of the present models. Here we show that the high-velocity deficit is due to the deficiency in close interactions of stars with the SMBH Sgr~$\rm{A}^{*}$, because an orbiting intermediate-mass black hole (IMBH) of about 15,000 Solar mass kicked away slowly approaching stars 50--250 million years ago. The SMBH-IMBH binary formed probably after the merger of the Milky Way with the Gaia-Sausage-Enceladus dwarf galaxy, and coalesced about 10 million years ago, leading to a gravitational recoil of Sgr~$\rm{A}^{*}$ at a velocity of 0.3--0.5$\, \rm{km\, s^{-1}}$ and to a change of the HVS ejection scenarios. The SMBH-IMBH binary scenario predicts the formation of the S-star cluster at the GC with the distribution of the orbital size and stellar ages that are well consistent with the observations.

astro-ph.HE↗

arXiv2025

Quantum ensemble learning with a programmable superconducting processor

Quantum machine learning is among the most exciting potential applications of quantum computing. However, the vulnerability of quantum information to environmental noises and the consequent high cost for realizing fault tolerance has impeded the quantum models from learning complex datasets. Here, we introduce AdaBoost.Q, a quantum adaptation of the classical adaptive boosting (AdaBoost) algorithm designed to enhance learning capabilities of quantum classifiers. Based on the probabilistic nature of quantum measurement, the algorithm improves the prediction accuracy by refining the attention mechanism during the adaptive training and combination of quantum classifiers. We experimentally demonstrate the versatility of our approach on a programmable superconducting processor, where we observe notable performance enhancements across various quantum machine learning models, including quantum neural networks and quantum convolutional neural networks. With AdaBoost.Q, we achieve an accuracy above 86% for a ten-class classification task over 10,000 test samples, and an accuracy of 100% for a quantum feature recognition task over 1,564 test samples. Our results demonstrate a foundational tool for advancing quantum machine learning towards practical applications, which has broad applicability to both the current noisy and the future fault-tolerant quantum devices.

quant-ph↗

arXiv2025

Active Learning from Scene Embeddings for End-to-End Autonomous Driving

In the field of autonomous driving, end-to-end deep learning models show great potential by learning driving decisions directly from sensor data. However, training these models requires large amounts of labeled data, which is time-consuming and expensive. Considering that the real-world driving data exhibits a long-tailed distribution where simple scenarios constitute a majority part of the data, we are thus inspired to identify the most challenging scenarios within it. Subsequently, we can efficiently improve the performance of the model by training with the selected data of the highest value. Prior research has focused on the selection of valuable data by empirically designed strategies. However, manually designed methods suffer from being less generalizable to new data distributions. Observing that the BEV (Bird's Eye View) features in end-to-end models contain all the information required to represent the scenario, we propose an active learning framework that relies on these vectorized scene-level features, called SEAD. The framework selects initial data based on driving-environmental information and incremental data based on BEV features. Experiments show that we only need 30\% of the nuScenes training data to achieve performance close to what can be achieved with the full dataset. The source code will be released.

cs.CV↗

arXiv2025

Exploring the Potential of Large Multimodal Models as Effective Alternatives for Pronunciation Assessment

Large Multimodal Models (LMMs) have demonstrated exceptional performance across a wide range of domains. This paper explores their potential in pronunciation assessment tasks, with a particular focus on evaluating the capabilities of the Generative Pre-trained Transformer (GPT) model, specifically GPT-4o. Our study investigates its ability to process speech and audio for pronunciation assessment across multiple levels of granularity and dimensions, with an emphasis on feedback generation and scoring. For our experiments, we use the publicly available Speechocean762 dataset. The evaluation focuses on two key aspects: multi-level scoring and the practicality of the generated feedback. Scoring results are compared against the manual scores provided in the Speechocean762 dataset, while feedback quality is assessed using Large Language Models (LLMs). The findings highlight the effectiveness of integrating LMMs with traditional methods for pronunciation assessment, offering insights into the model's strengths and identifying areas for further improvement.

cs.SD↗

arXiv2025

AttFC: Attention Fully-Connected Layer for Large-Scale Face Recognition with One GPU

Nowadays, with the advancement of deep neural networks (DNNs) and the availability of large-scale datasets, the face recognition (FR) model has achieved exceptional performance. However, since the parameter magnitude of the fully connected (FC) layer directly depends on the number of identities in the dataset. If training the FR model on large-scale datasets, the size of the model parameter will be excessively huge, leading to substantial demand for computational resources, such as time and memory. This paper proposes the attention fully connected (AttFC) layer, which could significantly reduce computational resources. AttFC employs an attention loader to generate the generative class center (GCC), and dynamically store the class center with Dynamic Class Container (DCC). DCC only stores a small subset of all class centers in FC, thus its parameter count is substantially less than the FC layer. Also, training face recognition models on large-scale datasets with one GPU often encounter out-of-memory (OOM) issues. AttFC overcomes this and achieves comparable performance to state-of-the-art methods.

cs.CV↗

arXiv2025

SplatPose: Geometry-Aware 6-DoF Pose Estimation from Single RGB Image via 3D Gaussian Splatting

6-DoF pose estimation is a fundamental task in computer vision with wide-ranging applications in augmented reality and robotics. Existing single RGB-based methods often compromise accuracy due to their reliance on initial pose estimates and susceptibility to rotational ambiguity, while approaches requiring depth sensors or multi-view setups incur significant deployment costs. To address these limitations, we introduce SplatPose, a novel framework that synergizes 3D Gaussian Splatting (3DGS) with a dual-branch neural architecture to achieve high-precision pose estimation using only a single RGB image. Central to our approach is the Dual-Attention Ray Scoring Network (DARS-Net), which innovatively decouples positional and angular alignment through geometry-domain attention mechanisms, explicitly modeling directional dependencies to mitigate rotational ambiguity. Additionally, a coarse-to-fine optimization pipeline progressively refines pose estimates by aligning dense 2D features between query images and 3DGS-synthesized views, effectively correcting feature misalignment and depth errors from sparse ray sampling. Experiments on three benchmark datasets demonstrate that SplatPose achieves state-of-the-art 6-DoF pose estimation accuracy in single RGB settings, rivaling approaches that depend on depth or multi-view images.

cs.CV↗

arXiv2025

LiNeS: Post-training Layer Scaling Prevents Forgetting and Enhances Model Merging

Fine-tuning pre-trained models has become the standard approach to endow them with specialized knowledge, but it poses fundamental challenges. In particular, \textit{(i)} fine-tuning often leads to catastrophic forgetting, where improvements on a target domain degrade generalization on other tasks, and \textit{(ii)} merging fine-tuned checkpoints from disparate tasks can lead to significant performance loss. To address these challenges, we introduce LiNeS, Layer-increasing Network Scaling, a post-training editing technique designed to preserve pre-trained generalization while enhancing fine-tuned task performance. LiNeS scales parameter updates linearly based on their layer depth within the network, maintaining shallow layers close to their pre-trained values to preserve general features while allowing deeper layers to retain task-specific representations. In multi-task model merging scenarios, layer-wise scaling of merged parameters reduces negative task interference. LiNeS demonstrates significant improvements in both single-task and multi-task settings across various benchmarks in vision and natural language processing. It mitigates forgetting, enhances out-of-distribution generalization, integrates seamlessly with existing multi-task model merging baselines improving their performance across benchmarks and model sizes, and can boost generalization when merging LLM policies aligned with different rewards via RLHF. Our method is simple to implement, computationally efficient and complementary to many existing techniques. Our source code is available at https://github.com/wang-kee/LiNeS

cs.LG↗

arXiv2025

SDPO: Segment-Level Direct Preference Optimization for Social Agents

Social agents powered by large language models (LLMs) can simulate human social behaviors but fall short in handling complex social dialogues. Direct Preference Optimization (DPO) has proven effective in aligning LLM behavior with human preferences across various agent tasks. However, standard DPO focuses solely on individual turns, which limits its effectiveness in multi-turn social interactions. Several DPO-based multi-turn alignment methods with session-level data have shown potential in addressing this problem.While these methods consider multiple turns across entire sessions, they are often overly coarse-grained, introducing training noise, and lack robust theoretical support. To resolve these limitations, we propose Segment-Level Direct Preference Optimization (SDPO), which dynamically select key segments within interactions to optimize multi-turn agent behavior. SDPO minimizes training noise and is grounded in a rigorous theoretical framework. Evaluations on the SOTOPIA benchmark demonstrate that SDPO-tuned agents consistently outperform both existing DPO-based methods and proprietary LLMs like GPT-4o, underscoring SDPO's potential to advance the social intelligence of LLM-based agents. We release our code and data at https://github.com/AlibabaResearch/DAMO-ConvAI/tree/main/SDPO.

cs.AI↗

arXiv2025

RIS-Assisted Received Adaptive Spatial Modulation for Wireless Communications

A novel wireless transmission scheme, as named the reconfigurable intelligent surface (RIS)-assisted received adaptive spatial modulation (RASM) scheme, is proposed in this paper. In this scheme, the adaptive spatial modulation (ASM)-based antennas selection works at the receiver by employing the characteristics of the RIS in each time slot, where the signal-to-noise ratio at specific selected antennas can be further enhanced with near few powers. Besides for the bits from constellation symbols, the extra bits can be mapped into the indices of receive antenna combinations and conveyed to the receiver through the ASM-based antenna-combination selection, thus providing higher spectral efficiency. To explicitly present the RASM scheme, the analytical performance of bit error rate of it is discussed in this paper. As a trade-off selection, the proposed scheme shows higher spectral efficiency and remains the satisfactory error performance. Simulation and analytical results demonstrate the better performance and exhibit more potential to apply in practical wireless communication.

cs.IT↗

arXiv2025

Model-independent Test of the Cosmic Anisotropy with Inverse Distance Ladder

If the Universe is endowed with cosmic anisotropy, it will have a preferred direction of expansion. Reconstructing the expansion history by Gaussian Process (GP) can be used to probe the cosmic anisotropy model-independently. In this paper, for the luminosity distance $d_L(z)$ reconstruction, we turn to the inverse distance ladder, where type Ia supernova (SNIa) from the Pantheon+ sample determine relative distances and strongly gravitationally lensed quasars from H0LiCOW sample anchor these relative distances with some absolute distance measurements. By isolating the anisotropic information that could be carried by the Hubble constant $H_0$ and obtaining constraints on the intrinsic parameter of SNIa, the absolute magnitude $M=-19.2522^{+0.0594}_{-0.0649}$ (at $68\%$ CL), we find that $d_L(z)$ reconstructions from samples located in different region of the Galactic coordinate system are almost consistent with each other, while only a very weak preference for the cosmic anisotropy is found.

astro-ph.CO↗

arXiv2025

RAG-Optimized Tibetan Tourism LLMs: Enhancing Accuracy and Personalization

With the development of the modern social economy, tourism has become an important way to meet people's spiritual needs, bringing development opportunities to the tourism industry. However, existing large language models (LLMs) face challenges in personalized recommendation capabilities and the generation of content that can sometimes produce hallucinations. This study proposes an optimization scheme for Tibet tourism LLMs based on retrieval-augmented generation (RAG) technology. By constructing a database of tourist viewpoints and processing the data using vectorization techniques, we have significantly improved retrieval accuracy. The application of RAG technology effectively addresses the hallucination problem in content generation. The optimized model shows significant improvements in fluency, accuracy, and relevance of content generation. This research demonstrates the potential of RAG technology in the standardization of cultural tourism information and data analysis, providing theoretical and technical support for the development of intelligent cultural tourism service systems.

cs.CL↗

arXiv2025

SLVR: Securely Leveraging Client Validation for Robust Federated Learning

Federated Learning (FL) enables collaborative model training while keeping client data private. However, exposing individual client updates makes FL vulnerable to reconstruction attacks. Secure aggregation mitigates such privacy risks but prevents the server from verifying the validity of each client update, creating a privacy-robustness tradeoff. Recent efforts attempt to address this tradeoff by enforcing checks on client updates using zero-knowledge proofs, but they support limited predicates and often depend on public validation data. We propose SLVR, a general framework that securely leverages clients' private data through secure multi-party computation. By utilizing clients' data, SLVR not only eliminates the need for public validation data, but also enables a wider range of checks for robustness, including cross-client accuracy validation. It also adapts naturally to distribution shifts in client data as it can securely refresh its validation data up-to-date. Our empirical evaluations show that SLVR improves robustness against model poisoning attacks, particularly outperforming existing methods by up to 50% under adaptive attacks. Additionally, SLVR demonstrates effective adaptability and stable convergence under various distribution shift scenarios.

cs.CR↗

arXiv2025

Bidirectional Piggybacking Design for Systematic Nodes with Sub-Packetization $l=2$

In 2013, Rashmi et al. proposed the piggybacking design framework to reduce the repair bandwidth of $(n,k;l)$ MDS array codes with small sub-packetization $l$ and it has been studied extensively in recent years. In this work, we propose an explicit bidirectional piggybacking design (BPD) with sub-packetization $l=2$ and the field size $q=O(n^{\lfloor r/2 \rfloor \!+\!1})$ for systematic nodes, where $r=n-k$ equals the redundancy of an $(n,k)$ linear code. And BPD has lower average repair bandwidth than previous piggybacking designs for $l=2$ when $r\geq 3$. Surprisingly, we can prove that the field size $q\leq 256$ is sufficient when $n\leq 15$ and $n-k\leq 4$. For example, we provide the BPD for the $(14,10)$ Reed-Solomon (RS) code over $\mathbb{F}_{2^8}$ and obtain approximately $41\%$ savings in the average repair bandwidth for systematic nodes compared with the trivial repair approach. This is the lowest repair bandwidth achieved so far for $(14,10)_{256}$ RS codes with sub-packetization $l=2$.

cs.IT↗

arXiv2025

Spectral Sufficient Conditions for Graph Factors

The $\{K_{1,1}, K_{1,2},C_m: m\geq3\}$-factor of a graph is a spanning subgraph whose each component is an element of $\{K_{1,1}, K_{1,2},C_m: m\geq3\}$. In this paper, through the graph spectral methods, we establish the lower bound of the signless Laplacian spectral radius and the upper bound of the distance spectral radius to determine whether a graph admits a $\{K_2\}$-factor. We get a lower bound on the size (resp. the spectral radius) of $G$ to guarantee that $G$ contains a $\{K_{1,1}, K_{1,2},C_m: m\geq3\}$-factor. Then we determine an upper bound on the distance spectral radius of $G$ to ensure that $G$ has a $\{K_{1,1}, K_{1,2},C_m: m\geq3\}$-factor. Furthermore, by constructing extremal graphs, we show that the above all bounds are best possible.

math.CO↗