arXiv subjects

Cheng Zhang

Publications and source records attributed to Cheng Zhang.

At least 181 records · Page 10Linked to original sources

arXiv2025

LIRM: Large Inverse Rendering Model for Progressive Reconstruction of Shape, Materials and View-dependent Radiance Fields

We present Large Inverse Rendering Model (LIRM), a transformer architecture that jointly reconstructs high-quality shape, materials, and radiance fields with view-dependent effects in less than a second. Our model builds upon the recent Large Reconstruction Models (LRMs) that achieve state-of-the-art sparse-view reconstruction quality. However, existing LRMs struggle to reconstruct unseen parts accurately and cannot recover glossy appearance or generate relightable 3D contents that can be consumed by standard Graphics engines. To address these limitations, we make three key technical contributions to build a more practical multi-view 3D reconstruction framework. First, we introduce an update model that allows us to progressively add more input views to improve our reconstruction. Second, we propose a hexa-plane neural SDF representation to better recover detailed textures, geometry and material parameters. Third, we develop a novel neural directional-embedding mechanism to handle view-dependent effects. Trained on a large-scale shape and material dataset with a tailored coarse-to-fine training scheme, our model achieves compelling results. It compares favorably to optimization-based dense-view inverse rendering methods in terms of geometry and relighting accuracy, while requiring only a fraction of the inference time.

cs.CV↗

arXiv2025

Integrating Learning-Based Manipulation and Physics-Based Locomotion for Whole-Body Badminton Robot Control

Learning-based methods, such as imitation learning (IL) and reinforcement learning (RL), can produce excel control policies over challenging agile robot tasks, such as sports robot. However, no existing work has harmonized learning-based policy with model-based methods to reduce training complexity and ensure the safety and stability for agile badminton robot control. In this paper, we introduce Hamlet, a novel hybrid control system for agile badminton robots. Specifically, we propose a model-based strategy for chassis locomotion which provides a base for arm policy. We introduce a physics-informed "IL+RL" training framework for learning-based arm policy. In this train framework, a model-based strategy with privileged information is used to guide arm policy training during both IL and RL phases. In addition, we train the critic model during IL phase to alleviate the performance drop issue when transitioning from IL to RL. We present results on our self-engineered badminton robot, achieving 94.5% success rate against the serving machine and 90.7% success rate against human players. Our system can be easily generalized to other agile mobile manipulation tasks such as agile catching and table tennis. Our project website: https://dreamstarring.github.io/HAMLET/.

cs.RO↗

arXiv2025

HalluLens: LLM Hallucination Benchmark

Large language models (LLMs) often generate responses that deviate from user input or training data, a phenomenon known as "hallucination." These hallucinations undermine user trust and hinder the adoption of generative AI systems. Addressing hallucinations is essential for the advancement of LLMs. This paper introduces a comprehensive hallucination benchmark, incorporating both new extrinsic and existing intrinsic evaluation tasks, built upon clear taxonomy of hallucination. A major challenge in benchmarking hallucinations is the lack of a unified framework due to inconsistent definitions and categorizations. We disentangle LLM hallucination from "factuality," proposing a clear taxonomy that distinguishes between extrinsic and intrinsic hallucinations, to promote consistency and facilitate research. Extrinsic hallucinations, where the generated content is not consistent with the training data, are increasingly important as LLMs evolve. Our benchmark includes dynamic test set generation to mitigate data leakage and ensure robustness against such leakage. We also analyze existing benchmarks, highlighting their limitations and saturation. The work aims to: (1) establish a clear taxonomy of hallucinations, (2) introduce new extrinsic hallucination tasks, with data that can be dynamically regenerated to prevent saturation by leakage, (3) provide a comprehensive analysis of existing benchmarks, distinguishing them from factuality evaluations.

cs.CL↗

arXiv2025

Calibrating Verbal Uncertainty as a Linear Feature to Reduce Hallucinations

LLMs often adopt an assertive language style also when making false claims. Such ``overconfident hallucinations'' mislead users and erode trust. Achieving the ability to express in language the actual degree of uncertainty around a claim is therefore of great importance. We find that ``verbal uncertainty'' is governed by a single linear feature in the representation space of LLMs, and show that this has only moderate correlation with the actual ``semantic uncertainty'' of the model. We apply this insight and show that (1) the mismatch between semantic and verbal uncertainty is a better predictor of hallucinations than semantic uncertainty alone and (2) we can intervene on verbal uncertainty at inference time and reduce confident hallucinations on short-form answers, achieving an average relative reduction of ~30%.

cs.CL↗

arXiv2025

Single Document Image Highlight Removal via A Large-Scale Real-World Dataset and A Location-Aware Network

Reflective documents often suffer from specular highlights under ambient lighting, severely hindering text readability and degrading overall visual quality. Although recent deep learning methods show promise in highlight removal, they remain suboptimal for document images, primarily due to the lack of dedicated datasets and tailored architectural designs. To tackle these challenges, we present DocHR14K, a large-scale real-world dataset comprising 14,902 high-resolution image pairs across six document categories and various lighting conditions. To the best of our knowledge, this is the first high-resolution dataset for document highlight removal that captures a wide range of real-world lighting conditions. Additionally, motivated by the observation that the residual map between highlighted and clean images naturally reveals the spatial structure of highlight regions, we propose a simple yet effective Highlight Location Prior (HLP) to estimate highlight masks without human annotations. Building on this prior, we present the Location-Aware Laplacian Pyramid Highlight Removal Network (L2HRNet), which effectively removes highlights by leveraging estimated priors and incorporates diffusion module to restore details. Extensive experiments demonstrate that DocHR14K improves highlight removal under diverse lighting conditions. Our L2HRNet achieves state-of-the-art performance across three benchmark datasets, including a 5.01\% increase in PSNR and a 13.17\% reduction in RMSE on DocHR14K.

cs.CV↗

arXiv2025

Efficient hyperentanglement-based quantum secret sharing protocol

Quantum secret sharing (QSS) is a typical multipartite cryptographic primitive, which is an important part of quantum communication network. Existing QSS protocols generally require basis selection and matching, which would increase the quantum resource consumption and classical communication round, and also face weak random security vulnerabilities. We propose an efficient hyperentanglement-based QSS protocol without the basis selection, in which the dealer and partners share a polarization-momentum hyperentangled Greenberger-Horne-Zeilinger (GHZ) state and encode keys in the polarization degree of freedom (DOF). The dealer decodes the transmitted keys relying on the nonlocal hyperentanglement-assisted polarization GHZ state analysis. Our QSS protocol is unconditionally secure in theory. We develop simulation method to estimate its performance in practical environment. Compared with the QSS based on the GHZ state, our protocol has several advantages. First, it does not require basis selection and can completely distinguish eight polarization GHZ states, which can improve the utilization rate of entanglement resources to 100$\%$ and increase the key generation rate by an order of magnitude. Second, it only requires one round of classical communication, and thus can reduce key generation time by 69.4$\%$. Third, it can eliminate the weak random security vulnerabilities associated with the basis selection. Finally, our protocol only uses linear optical elements, which makes it practically feasible. Our QSS protocol has potential application in future quantum communication network.

quant-ph↗

arXiv2025

Vector encoding of phylogenetic trees by ordered leaf attachment

As part of work to connect phylogenetics with machine learning, there has been considerable recent interest in vector encodings of phylogenetic trees. We present a simple new "ordered leaf attachment" (OLA) method for uniquely encoding a binary, rooted phylogenetic tree topology as an integer vector. OLA encoding and decoding take linear time in the number of leaf nodes, and the set of vectors corresponding to trees is a simply-described subset of integer sequences. The OLA encoding is unique compared to other existing encodings in having these properties. The integer vector encoding induces a distance on the set of trees, and we investigate this distance in relation to the NNI and SPR distances.

q-bio.PE↗

arXiv2025

Panorama Generation From NFoV Image Done Right

Generating 360-degree panoramas from narrow field of view (NFoV) image is a promising computer vision task for Virtual Reality (VR) applications. Existing methods mostly assess the generated panoramas with InceptionNet or CLIP based metrics, which tend to perceive the image quality and is \textbf{not suitable for evaluating the distortion}. In this work, we first propose a distortion-specific CLIP, named Distort-CLIP to accurately evaluate the panorama distortion and discover the \textbf{``visual cheating''} phenomenon in previous works (\ie, tending to improve the visual results by sacrificing distortion accuracy). This phenomenon arises because prior methods employ a single network to learn the distinct panorama distortion and content completion at once, which leads the model to prioritize optimizing the latter. To address the phenomenon, we propose \textbf{PanoDecouple}, a decoupled diffusion model framework, which decouples the panorama generation into distortion guidance and content completion, aiming to generate panoramas with both accurate distortion and visual appeal. Specifically, we design a DistortNet for distortion guidance by imposing panorama-specific distortion prior and a modified condition registration mechanism; and a ContentNet for content completion by imposing perspective image information. Additionally, a distortion correction loss function with Distort-CLIP is introduced to constrain the distortion explicitly. The extensive experiments validate that PanoDecouple surpasses existing methods both in distortion and visual metrics.

cs.CV↗

arXiv2025

PanSplat: 4K Panorama Synthesis with Feed-Forward Gaussian Splatting

With the advent of portable 360° cameras, panorama has gained significant attention in applications like virtual reality (VR), virtual tours, robotics, and autonomous driving. As a result, wide-baseline panorama view synthesis has emerged as a vital task, where high resolution, fast inference, and memory efficiency are essential. Nevertheless, existing methods are typically constrained to lower resolutions (512 $\times$ 1024) due to demanding memory and computational requirements. In this paper, we present PanSplat, a generalizable, feed-forward approach that efficiently supports resolution up to 4K (2048 $\times$ 4096). Our approach features a tailored spherical 3D Gaussian pyramid with a Fibonacci lattice arrangement, enhancing image quality while reducing information redundancy. To accommodate the demands of high resolution, we propose a pipeline that integrates a hierarchical spherical cost volume and Gaussian heads with local operations, enabling two-step deferred backpropagation for memory-efficient training on a single A100 GPU. Experiments demonstrate that PanSplat achieves state-of-the-art results with superior efficiency and image quality across both synthetic and real-world datasets. Code is available at https://github.com/chengzhag/PanSplat.

cs.CV↗

arXiv2025

Semi-Implicit Functional Gradient Flow for Efficient Sampling

Particle-based variational inference methods (ParVIs) use nonparametric variational families represented by particles to approximate the target distribution according to the kernelized Wasserstein gradient flow for the Kullback-Leibler (KL) divergence. Although functional gradient flows have been introduced to expand the kernel space for better flexibility, the deterministic updating mechanism may limit exploration and require expensive repetitive runs for new samples. In this paper, we propose Semi-Implicit Functional Gradient flow (SIFG), a functional gradient ParVI method that uses perturbed particles with Gaussian noise as the approximation family. We show that the corresponding functional gradient flow, which can be estimated via denoising score matching with neural networks, exhibits strong theoretical convergence guarantees due to a higher-order smoothness brought to the approximation family via Gaussian perturbation. In addition, we present an adaptive version of our method that automatically selects the appropriate noise magnitude during sampling, striking a good balance between exploration efficiency and approximation accuracy. Extensive experiments on both simulated and real-world datasets demonstrate the effectiveness and efficiency of the proposed framework.

stat.ML↗

arXiv2025

GarmentCrafter: Progressive Novel View Synthesis for Single-View 3D Garment Reconstruction and Editing

We introduce GarmentCrafter, a new approach that enables non-professional users to create and modify 3D garments from a single-view image. While recent advances in image generation have facilitated 2D garment design, creating and editing 3D garments remains challenging for non-professional users. Existing methods for single-view 3D reconstruction often rely on pre-trained generative models to synthesize novel views conditioning on the reference image and camera pose, yet they lack cross-view consistency, failing to capture the internal relationships across different views. In this paper, we tackle this challenge through progressive depth prediction and image warping to approximate novel views. Subsequently, we train a multi-view diffusion model to complete occluded and unknown clothing regions, informed by the evolving camera pose. By jointly inferring RGB and depth, GarmentCrafter enforces inter-view coherence and reconstructs precise geometries and fine details. Extensive experiments demonstrate that our method achieves superior visual fidelity and inter-view coherence compared to state-of-the-art single-view 3D garment reconstruction methods.

cs.GR↗

arXiv2025

Locking Machine Learning Models into Hardware

Modern machine learning (ML) models are expensive IP and business competitiveness often depends on keeping this IP confidential. This in turn restricts how these models are deployed; for example, it is unclear how to deploy a model on-device without inevitably leaking the underlying model. At the same time, confidential computing technologies such as multi-party computation or homomorphic encryption remain impractical for wide adoption. In this paper, we take a different approach and investigate the feasibility of ML-specific mechanisms that deter unauthorized model use by restricting the model to only be usable on specific hardware, making adoption on unauthorized hardware inconvenient. That way, even if IP is compromised, it cannot be trivially used without specialised hardware or major model adjustment. In a sense, we seek to enable cheap \emph{locking of machine learning models into specific hardware}. We demonstrate that \emph{locking} mechanisms are feasible by either targeting efficiency of model representations, making such models incompatible with quantization, or tying the model's operation to specific characteristics of hardware, such as the number of clock cycles for arithmetic operations. We demonstrate that locking comes with negligible overheads, while significantly restricting usability of the resultant model on unauthorized hardware.

cs.CR↗

arXiv2025

HVI: A New Color Space for Low-light Image Enhancement

Low-Light Image Enhancement (LLIE) is a crucial computer vision task that aims to restore detailed visual information from corrupted low-light images. Many existing LLIE methods are based on standard RGB (sRGB) space, which often produce color bias and brightness artifacts due to inherent high color sensitivity in sRGB. While converting the images using Hue, Saturation and Value (HSV) color space helps resolve the brightness issue, it introduces significant red and black noise artifacts. To address this issue, we propose a new color space for LLIE, namely Horizontal/Vertical-Intensity (HVI), defined by polarized HS maps and learnable intensity. The former enforces small distances for red coordinates to remove the red artifacts, while the latter compresses the low-light regions to remove the black artifacts. To fully leverage the chromatic and intensity information, a novel Color and Intensity Decoupling Network (CIDNet) is further introduced to learn accurate photometric mapping function under different lighting conditions in the HVI space. Comprehensive results from benchmark and ablation experiments show that the proposed HVI color space with CIDNet outperforms the state-of-the-art methods on 10 datasets. The code is available at https://github.com/Fediory/HVI-CIDNet.

cs.CV↗

arXiv2025

The elliptic lattice KdV system revisited

In a previous paper [Nijhoff,Puttock,2003], a 2-parameter extension of the lattice potential KdV equation was derived, associated with an elliptic curve. This comprises a rather complicated 3-component system on the quad lattice which contains the moduli of the elliptic curve as parameters. In the present paper, we investigate this system further and, among other results, we derive a 2-component multiquartic form of the system on the quad lattice. Furthermore, we construct an elliptic Yang-Baxter map, and study the associated continuous and semi-discrete systems. In particular, we derive the so-called ``generating PDE'' for this system, comprising a 6-component system of second order PDEs which could be considered to constitute an elliptic extension of the Ernst equations of General Relativity.

nlin.SI↗

arXiv2025

QERA: an Analytical Framework for Quantization Error Reconstruction

The growing number of parameters and computational demands of large language models (LLMs) present significant challenges for their efficient deployment. Recently, there is an increasing interest in quantizing weights to extremely low precision while offsetting the resulting error with low-rank, high-precision error reconstruction terms. The combination of quantization and low-rank approximation is now popular in both adapter-based, parameter-efficient fine-tuning methods such as LoftQ and low-precision inference techniques including ZeroQuant-V2. Usually, the low-rank terms are calculated via the singular value decomposition (SVD) of the weight quantization error, minimizing the Frobenius and spectral norms of the weight approximation error. Recent methods like LQ-LoRA and LQER introduced hand-crafted heuristics to minimize errors in layer outputs (activations) rather than weights, resulting improved quantization results. However, these heuristic methods lack an analytical solution to guide the design of quantization error reconstruction terms. In this paper, we revisit this problem and formulate an analytical framework, named Quantization Error Reconstruction Analysis (QERA), and offer a closed-form solution to the problem. We show QERA benefits both existing low-precision fine-tuning and inference methods -- QERA achieves a fine-tuned accuracy gain of $Δ_{\text{acc}}$ = 6.05% of 2-bit RoBERTa-base on GLUE compared to LoftQ; and obtains $Δ_{\text{acc}}$ = 2.97% higher post-training quantization accuracy of 4-bit Llama-3.1-70B on average than ZeroQuant-V2 and $Δ_{\text{ppl}}$ = - 0.28 lower perplexity on WikiText2 than LQER.

cs.LG↗

arXiv2025

Coupled hydro-aero-turbo dynamics of liquid-tank system for wave energy harvesting: Numerical modellings and scaled prototype tests

An integrated numerical model is proposed for the first time to explore the coupled hydro-aero-turbo dynamics of wave-energy-harvesting (WEH) liquid tanks. A scaled prototype of the WEH liquid tank with an impulse air turbine system is made to experimentally validate the numerical model.Multi-layered impulse air turbine systems (MLATS) are creatively introduced into the liquid-tank system. The inherent mechanisms of the coupled hydro-aero-turbo dynamics of the WEH liquid tank with different turbine properties are systematically investigated.Compared with the experimental data, the numerical model can accurately reproduce the rotor speed, liquid motion, and air pressure of the WEH liquid tank. Upon analysing mechanical parameters of the turbine rotor, it is found that the rotor's moment of inertia mainly affects the rotor speed's variation range, while the damping coefficient significantly influences the averaged rotor speed. The optimal power take-off damping for the WEH liquid tank is identified. Considering the efficiency performances of three MLATSs, improving Turbine-L1 to Turbine-L2 or Turbine-L3 can increase the averaged power output by about 25% or 40%, respectively.Increasing the tank breadth can effectively boost the power output in a nonlinear way.Under the considered excitation conditions, if the tank breadth is doubled, the maximum averaged power output can be increased by around four times. Through a series of failure tests, Turbine-L3 shows greater reliability in extreme conditions compared to a conventional single-rotor turbine. Even if the most important rotor of Turbine-L3 fails to work, the maximum loss of the averaged power output is only 44%. The present WEH liquid with Turbine - L3 shows improved efficiency and reliability compared to the conventional liquid-tank system with a single-rotor turbine.

physics.flu-dyn↗

arXiv2025

SpellRing: Recognizing Continuous Fingerspelling in American Sign Language using a Ring

Fingerspelling is a critical part of American Sign Language (ASL) recognition and has become an accessible optional text entry method for Deaf and Hard of Hearing (DHH) individuals. In this paper, we introduce SpellRing, a single smart ring worn on the thumb that recognizes words continuously fingerspelled in ASL. SpellRing uses active acoustic sensing (via a microphone and speaker) and an inertial measurement unit (IMU) to track handshape and movement, which are processed through a deep learning algorithm using Connectionist Temporal Classification (CTC) loss. We evaluated the system with 20 ASL signers (13 fluent and 7 learners), using the MacKenzie-Soukoref Phrase Set of 1,164 words and 100 phrases. Offline evaluation yielded top-1 and top-5 word recognition accuracies of 82.45% (9.67%) and 92.42% (5.70%), respectively. In real-time, the system achieved a word error rate (WER) of 0.099 (0.039) on the phrases. Based on these results, we discuss key lessons and design implications for future minimally obtrusive ASL recognition wearables.

cs.HC↗

arXiv2025

Unexpected large electrostatic gating by pyroelectric charge accumulation

Pyroelectricity refers to the accumulation of charges due to changes in the spontaneous polarization of ferroelectric materials when subjected to temperature variations. Typically, these pyroelectric charges are considered unstable and dissipate quickly through interactions with the external environment. Consequently, the pyroelectric effect has been largely overlooked in ferroelectric field-effect transistors. In this work, we leverage the van der Waals interface of hBN to achieve a substantial and long-term electrostatic gating effect in graphene devices via the pyroelectric properties of a ferroelectric LiNbO3 substrate. Upon cooling, the polarization change in LiNbO3 induces high doping concentrations up to 1013 cm-2 in the adjacent graphene. Through a combination of transport measurements and non-contact techniques, we demonstrate that the pyroelectric charge accumulation, as well as its enhancement in electric fields, are responsible for this unexpectedly high doping level. Our findings introduce a novel mechanism for voltage-free electrostatic gating control with long retention.

cond-mat.mtrl-sci↗