Searcharxiv⌕ Search

arXiv subjects

Cheng Zhao

Publications and source records attributed to Cheng Zhao.

At least 37 records · Page 2Linked to original sources

Tracking-Guided 4D Generation: Foundation-Tracker Motion Priors for 3D Model Animation

Generating dynamic 4D objects from sparse inputs is difficult because it demands joint preservation of appearance and motion coherence across views and time while suppressing artifacts and temporal drift. We hypothesize that the view discrepancy arises from supervision limited to pixel- or latent-space video-diffusion losses, which lack explicitly temporally aware, feature-level tracking guidance. We present \emph{Track4DGen}, a two-stage framework that couples a multi-view video diffusion model with a foundation point tracker and a hybrid 4D Gaussian Splatting (4D-GS) reconstructor. The central idea is to explicitly inject tracker-derived motion priors into intermediate feature representations for both multi-view video generation and 4D-GS. In Stage One, we enforce dense, feature-level point correspondences inside the diffusion generator, producing temporally consistent features that curb appearance drift and enhance cross-view coherence. In Stage Two, we reconstruct a dynamic 4D-GS using a hybrid motion encoding that concatenates co-located diffusion features (carrying Stage-One tracking priors) with Hex-plane features, and augment them with 4D Spherical Harmonics for higher-fidelity dynamics modeling. \emph{Track4DGen} surpasses baselines on both multi-view video generation and 4D generation benchmarks, yielding temporally stable, text-editable 4D assets. Lastly, we curate \emph{Sketchfab28}, a high-quality dataset for benchmarking object-centric 4D generation and fostering future research.

cs.CV↗

Theory and Design of Extended PID Control for Stochastic Systems with Structural Uncertainties

Since the classical proportional-integral-derivative (PID) controller has continued to be the most widely used feedback methods in engineering systems by far, it is crucial to investigate the working mechanism of PID in dealing with nonlinearity, uncertainty and random noises. Recently, Zhao and Guo (2022) has established the global stability of PID control for a class of uncertain nonlinear control systems with relative degree two without random perturbations. In this paper, we will consider a more general class of nonlinear stochastic systems with an arbitrary relative degree $n$, and discuss the stability and design of extended PID controller (a natural extension of PID). We demonstrate that, the closed-loop control systems will be globally stable in mean square with bounded tracking errors provided the extended PID parameters are selected from an $(n+1)$-dimensional unbounded set, even if both the system nonlinear drift and diffusion terms contain a wide range of structural uncertainties. Moreover, the steady-state tracking error is proved to be proportional to the noise intensity at the setpoint, which can also be made arbitrarily small by choosing the controller parameters suitably large.

math.OC↗

Scenes as Tokens: Multi-Scale Normal Distributions Transform Tokenizer for General 3D Vision-Language Understanding

Recent advances in 3D vision-language models (VLMs) highlight a strong potential for 3D scene understanding and reasoning. However, effectively tokenizing 3D scenes into holistic scene tokens, and leveraging these tokens across diverse 3D understanding tasks, remain highly challenging. We present NDTokenizer3D, a generalist 3D VLM that performs a wide range of 3D scene understanding tasks while naturally supporting human interactions, thereby bridging language-level reasoning with 3D spatial understanding. The core of our approach is a novel three-stage scene tokenization pipeline built upon a Multi-Scale Normal Distributions Transform (NDT) representation, paired with a Multi-Scale NDT Decoder (MSDec). Specifically, NDTokenizer3D first constructs a multi-scale NDT representation from raw high-resolution point clouds, preserving both global context and fine-grained geometric details. Next, the MSDec progressively fuses cross-scale NDT features, producing holistic scene tokens consumable by LLM endpoints. Beyond tokenization, MSDec is repurposed as a general interface for human-interactive prompting (points, boxes, masks) and segmentation-mask decoding, unifying diverse 3D scene understanding tasks within a single architecture. With this compact and unified design, NDTokenizer3D offers a fine-grained, general-purpose 3D VLM, achieving remarkable improvements in 3D Referring Segmentation, 3D Visual Question Answering, and 3D Dense Captioning.

cs.CV↗

Dynamical Dark Energy in light of the DESI DR2 Baryonic Acoustic Oscillations Measurements

Understanding whether cosmic acceleration arises from a cosmological constant or a dynamical component is a central goal of cosmology, and the Dark Energy Spectroscopic Instrument (DESI) enables stringent tests with high-precision distance measurements. We analyze baryon acoustic oscillation (BAO) measurements from DESI Data Release 1 (DR1) and Data Release 2 (DR2), combined with Type Ia supernovae and a cosmic microwave background (CMB) distance prior. With the larger statistical power and wider redshift coverage of DR2, the preference for dynamical dark energy does not diminish relative to DR1. Using both a shape-function reconstruction and non-parametric approaches with a Horndeski-motivated correlation prior, we find that the dark-energy equation of state $w(z)$ varies with redshift. BAO data alone yield modest constraints, but in combination with independent supernova compilations and the CMB prior they strengthen the evidence for dynamics. Bayesian model comparison shows moderate support for departures from $Λ$CDM when multiple degrees of freedom in $w(z)$ are allowed, corresponding to $\approx3σ$ tension with $Λ$CDM (and higher for some data sets). Despite methodological differences, our results are consistent with companion DESI papers, underscoring the complementarity of approaches. Possible systematics remain under study; forthcoming DESI, \emph{Euclid}, and next-generation CMB data will provide decisive tests.

astro-ph.CO↗

DeCafNet: Delegate and Conquer for Efficient Temporal Grounding in Long Videos

Long Video Temporal Grounding (LVTG) aims at identifying specific moments within lengthy videos based on user-provided text queries for effective content retrieval. The approach taken by existing methods of dividing video into clips and processing each clip via a full-scale expert encoder is challenging to scale due to prohibitive computational costs of processing a large number of clips in long videos. To address this issue, we introduce DeCafNet, an approach employing ``delegate-and-conquer'' strategy to achieve computation efficiency without sacrificing grounding performance. DeCafNet introduces a sidekick encoder that performs dense feature extraction over all video clips in a resource-efficient manner, while generating a saliency map to identify the most relevant clips for full processing by the expert encoder. To effectively leverage features from sidekick and expert encoders that exist at different temporal resolutions, we introduce DeCaf-Grounder, which unifies and refines them via query-aware temporal aggregation and multi-scale temporal refinement for accurate grounding. Experiments on two LTVG benchmark datasets demonstrate that DeCafNet reduces computation by up to 47\% while still outperforming existing methods, establishing a new state-of-the-art for LTVG in terms of both efficiency and performance. Our code is available at https://github.com/ZijiaLewisLu/CVPR2025-DeCafNet.

cs.CV↗

Cosmological Constraints with Void Lensing I: the Simulation-Based Inference Framework

We present a Simulation-Based Inference (SBI) framework for cosmological parameter estimation via void lensing analysis. Despite the absence of an analytical model of void lensing, SBI can effectively learn posterior distributions through forward modeling of mock data. We develop a forward modeling pipeline that accounts for both cosmology and the galaxy-halo connection. By training a neural density estimator on simulated data, we infer the posteriors of two cosmological parameters, $Ω_m$ and $S_8$. Validation tests are conducted on posteriors derived from different cosmological parameters and a fiducial sample. The results demonstrate that SBI provides unbiased estimates of mean values and accurate uncertainties. These findings highlight the potential to apply void lensing analysis to observational data even without an analytical void lensing model.

astro-ph.CO↗

SplatFlow: Self-Supervised Dynamic Gaussian Splatting in Neural Motion Flow Field for Autonomous Driving

Most existing Dynamic Gaussian Splatting methods for complex dynamic urban scenarios rely on accurate object-level supervision from expensive manual labeling, limiting their scalability in real-world applications. In this paper, we introduce SplatFlow, a Self-Supervised Dynamic Gaussian Splatting within Neural Motion Flow Fields (NMFF) to learn 4D space-time representations without requiring tracked 3D bounding boxes, enabling accurate dynamic scene reconstruction and novel view RGB/depth/flow synthesis. SplatFlow designs a unified framework to seamlessly integrate time-dependent 4D Gaussian representation within NMFF, where NMFF is a set of implicit functions to model temporal motions of both LiDAR points and Gaussians as continuous motion flow fields. Leveraging NMFF, SplatFlow effectively decomposes static background and dynamic objects, representing them with 3D and 4D Gaussian primitives, respectively. NMFF also models the correspondences of each 4D Gaussian across time, which aggregates temporal features to enhance cross-view consistency of dynamic components. SplatFlow further improves dynamic object identification by distilling features from 2D foundation models into 4D space-time representation. Comprehensive evaluations conducted on the Waymo and KITTI Datasets validate SplatFlow's state-of-the-art (SOTA) performance for both image reconstruction and novel view synthesis in dynamic urban scenarios.

cs.CV↗

Measuring the Mean Free Path of HI Ionizing Photons at $3.2\leq z\leq4.6$ with DESI Y1 Quasars

The mean free path of ionizing photons ($λ_\mathrm{mfp}^{912}$) in the intergalactic medium (IGM) is a crucial quantity in modelling the ionization state of IGM and the extragalactic ultraviolet background (EUVB), and is widely used in hydrodynamical simulations of galaxies and reionization. We construct the largest quasar spectrum dataset to date -- 12,595 $\mathrm{S/N}>3$ spectra -- using the Y1 observation of Dark Energy Spectroscopic Instrument (DESI) to make the most precise model-independent measurement of the mean free path at $3.2\leq z\leq 4.6$. By stacking the spectra in 17 redshift bins and modelling the Lyman continuum profile, we get a redshift evolution $λ_\mathrm{mfp}^{912}\propto(1+z)^{-4.27}$ at $2\leq z\leq 5$, which is much shallower than previous estimate $λ_\mathrm{mfp}^{912}\propto(1+z)^{-5.4}$. We then explore the sources of systematic bias, including the choice of intrinsic quasar continuum, the consideration of Lyman series opacity and Lyman limit opacity evolution and the definition of $λ_\mathrm{mfp}^{912}$. Combining our results with estimates of $λ_\mathrm{mfp}^{912}$ at higher redshifts, we conclude at high confidence that the evolution in $λ_\mathrm{mfp}^{912}$ steepens at $z \approx 5$. We interpret this inflection as the transition from the end of HI reionization to a fully ionized plasma which characterizes the intergalactic medium of the past $\sim10$ billion years.

astro-ph.CO↗

ELG Spectroscopic Systematics Analysis of the DESI Data Release 1

Dark Energy Spectroscopic Instrument (DESI) uses more than 2.4 million Emission Line Galaxies (ELGs) for 3D large-scale structure (LSS) analyses in its Data Release 1 (DR1). Such large statistics enable thorough research on systematic uncertainties. In this study, we focus on spectroscopic systematics of ELGs. The redshift success rate ($f_{\rm goodz}$) is the relative fraction of secure redshifts among all measurements. It depends on observing conditions, thus introduces non-cosmological variations to the LSS. We, therefore, develop the redshift failure weight ($w_{\rm zfail}$) and a per-fibre correction ($η_{\rm zfail}$) to mitigate these dependences. They have minor influences on the galaxy clustering. For ELGs with a secure redshift, there are two subtypes of systematics: 1) catastrophics (large) that only occur in a few samples; 2) redshift uncertainty (small) that exists for all samples. The catastrophics represent 0.26\% of the total DR1 ELGs, composed of the confusion between O\,\textsc{ii} and sky residuals, double objects, total catastrophics and others. We simulate the realistic 0.26\% catastrophics of DR1 ELGs, the hypothetical 1\% catastrophics, and the truncation of the contaminated $1.31<z<1.33$ in the \textsc{AbacusSummit} ELG mocks. Their $P_\ell$ show non-negligible bias from the uncontaminated mocks. But their influences on the redshift space distortions (RSD) parameters are smaller than $0.2σ$. The redshift uncertainty of \Yone ELGs is 8.5 km/s with a Lorentzian profile. The code for implementing the catastrophics and redshift uncertainty on mocks can be found in https://github.com/Jiaxi-Yu/modelling_spectro_sys.

astro-ph.CO↗

Optimal Adaptive Control of Linear Stochastic Systems with Quadratic Cost Function

In this paper, we consider the adaptive linear quadratic Gaussian control problem, where both the linear transformation matrix of the state $A$ and the control gain matrix $B$ are unknown. The proposed adaptive optimal control only assumes that $(A, B)$ is stabilizable and $(A, Q^{1/2})$ is detectable, where $Q$ is the weighting matrix of the state in the quadratic cost function. This condition significantly weakens the classic assumptions used in the literature. To tackle this problem, a weighted least squares algorithm is modified by using random regularization method, which can ensure uniform stabilizability and uniform detectability of the family of estimated models. At the same time, a diminishing excitation is incorporated into the design of the proposed adaptive control to guarantee strong consistency of the desired components of the estimates. Finally, by utilizing this family of estimates, even if not all components of them converge to the true values, it is demonstrated that a certainty equivalence control with such a diminishing excitation is optimal for an ergodic quadratic cost function.

math.OC↗

Understanding Literary Texts by LLMs: A Case Study of Ancient Chinese Poetry

The birth and rapid development of large language models (LLMs) have caused quite a stir in the field of literature. Once considered unattainable, AI's role in literary creation is increasingly becoming a reality. In genres such as poetry, jokes, and short stories, numerous AI tools have emerged, offering refreshing new perspectives. However, it's difficult to further improve the quality of these works. This is primarily because understanding and appreciating a good literary work involves a considerable threshold, such as knowledge of literary theory, aesthetic sensibility, interdisciplinary knowledge. Therefore, authoritative data in this area is quite lacking. Additionally, evaluating literary works is often complex and hard to fully quantify, which directly hinders the further development of AI creation. To address this issue, this paper attempts to explore the mysteries of literary texts from the perspective of LLMs, using ancient Chinese poetry as an example for experimentation. First, we collected a variety of ancient poems from different sources and had experts annotate a small portion of them. Then, we designed a range of comprehension metrics based on LLMs to evaluate all these poems. Finally, we analyzed the correlations and differences between various poem collections to identify literary patterns. Through our experiments, we observed a series of enlightening phenomena that provide technical support for the future development of high-level literary creation based on LLMs.

cs.CL↗

AdaOcc: Adaptive-Resolution Occupancy Prediction

Autonomous driving in complex urban scenarios requires 3D perception to be both comprehensive and precise. Traditional 3D perception methods focus on object detection, resulting in sparse representations that lack environmental detail. Recent approaches estimate 3D occupancy around vehicles for a more comprehensive scene representation. However, dense 3D occupancy prediction increases computational demands, challenging the balance between efficiency and resolution. High-resolution occupancy grids offer accuracy but demand substantial computational resources, while low-resolution grids are efficient but lack detail. To address this dilemma, we introduce AdaOcc, a novel adaptive-resolution, multi-modal prediction approach. Our method integrates object-centric 3D reconstruction and holistic occupancy prediction within a single framework, performing highly detailed and precise 3D reconstruction only in regions of interest (ROIs). These high-detailed 3D surfaces are represented in point clouds, thus their precision is not constrained by the predefined grid resolution of the occupancy map. We conducted comprehensive experiments on the nuScenes dataset, demonstrating significant improvements over existing methods. In close-range scenarios, we surpass previous baselines by over 13% in IOU, and over 40% in Hausdorff distance. In summary, AdaOcc offers a more versatile and effective framework for delivering accurate 3D semantic occupancy prediction across diverse driving scenarios.

cs.CV↗

SUP-NeRF: A Streamlined Unification of Pose Estimation and NeRF for Monocular 3D Object Reconstruction

Monocular 3D reconstruction for categorical objects heavily relies on accurately perceiving each object's pose. While gradient-based optimization in a NeRF framework updates the initial pose, this paper highlights that scale-depth ambiguity in monocular object reconstruction causes failures when the initial pose deviates moderately from the true pose. Consequently, existing methods often depend on a third-party 3D object to provide an initial object pose, leading to increased complexity and generalization issues. To address these challenges, we present SUP-NeRF, a Streamlined Unification of object Pose estimation and NeRF-based object reconstruction. SUP-NeRF decouples the object's dimension estimation and pose refinement to resolve the scale-depth ambiguity, and introduces a camera-invariant projected-box representation that generalizes cross different domains. While using a dedicated pose estimator that smoothly integrates into an object-centric NeRF, SUP-NeRF is free from external 3D detectors. SUP-NeRF achieves state-of-the-art results in both reconstruction and pose estimation tasks on the nuScenes dataset. Furthermore, SUP-NeRF exhibits exceptional cross-dataset generalization on the KITTI and Waymo datasets, surpassing prior methods with up to 50\% reduction in rotation and translation error.

cs.CV↗

TCLC-GS: Tightly Coupled LiDAR-Camera Gaussian Splatting for Autonomous Driving

Most 3D Gaussian Splatting (3D-GS) based methods for urban scenes initialize 3D Gaussians directly with 3D LiDAR points, which not only underutilizes LiDAR data capabilities but also overlooks the potential advantages of fusing LiDAR with camera data. In this paper, we design a novel tightly coupled LiDAR-Camera Gaussian Splatting (TCLC-GS) to fully leverage the combined strengths of both LiDAR and camera sensors, enabling rapid, high-quality 3D reconstruction and novel view RGB/depth synthesis. TCLC-GS designs a hybrid explicit (colorized 3D mesh) and implicit (hierarchical octree feature) 3D representation derived from LiDAR-camera data, to enrich the properties of 3D Gaussians for splatting. 3D Gaussian's properties are not only initialized in alignment with the 3D mesh which provides more completed 3D shape and color information, but are also endowed with broader contextual information through retrieved octree implicit features. During the Gaussian Splatting optimization process, the 3D mesh offers dense depth information as supervision, which enhances the training process by learning of a robust geometry. Comprehensive evaluations conducted on the Waymo Open Dataset and nuScenes Dataset validate our method's state-of-the-art (SOTA) performance. Utilizing a single NVIDIA RTX 3090 Ti, our method demonstrates fast training and achieves real-time RGB and depth rendering at 90 FPS in resolution of 1920x1280 (Waymo), and 120 FPS in resolution of 1600x900 (nuScenes) in urban scenarios.

cs.CV↗

The Completed SDSS-IV extended Baryon Oscillation Spectroscopic Survey: Cosmological Implications from two Decades of Spectroscopic Surveys at the Apache Point observatory

We present the cosmological implications from final measurements of clustering using galaxies, quasars, and Ly$α$ forests from the completed Sloan Digital Sky Survey (SDSS) lineage of experiments in large-scale structure. These experiments, composed of data from SDSS, SDSS-II, BOSS, and eBOSS, offer independent measurements of baryon acoustic oscillation (BAO) measurements of angular-diameter distances and Hubble distances relative to the sound horizon, $r_d$, from eight different samples and six measurements of the growth rate parameter, $fσ_8$, from redshift-space distortions (RSD). This composite sample is the most constraining of its kind and allows us to perform a comprehensive assessment of the cosmological model after two decades of dedicated spectroscopic observation. We show that the BAO data alone are able to rule out dark-energy-free models at more than eight standard deviations in an extension to the flat, $Λ$CDM model that allows for curvature. When combined with Planck Cosmic Microwave Background (CMB) measurements of temperature and polarization the BAO data provide nearly an order of magnitude improvement on curvature constraints. The RSD measurements indicate a growth rate that is consistent with predictions from Planck primary data and with General Relativity. When combining the results of SDSS BAO and RSD with external data, all multiple-parameter extensions remain consistent with a $Λ$CDM model. Regardless of cosmological model, the precision on $Ω_Λ$, $H_0$, and $σ_8$, remains at roughly 1\%, showing changes of less than 0.6\% in the central values between models. The inverse distance ladder measurement under a o$w_0w_a$CDM yields $H_0= 68.20 \pm 0.81 \, \rm km\, s^{-1} Mpc^{-1}$, remaining in tension with several direct determination methods. (abridged)

astro-ph.CO↗

FlowWalker: A Memory-efficient and High-performance GPU-based Dynamic Graph Random Walk Framework

Dynamic graph random walk (DGRW) emerges as a practical tool for capturing structural relations within a graph. Effectively executing DGRW on GPU presents certain challenges. First, existing sampling methods demand a pre-processing buffer, causing substantial space complexity. Moreover, the power-law distribution of graph vertex degrees introduces workload imbalance issues, rendering DGRW embarrassed to parallelize. In this paper, we propose FlowWalker, a GPU-based dynamic graph random walk framework. FlowWalker implements an efficient parallel sampling method to fully exploit the GPU parallelism and reduce space complexity. Moreover, it employs a sampler-centric paradigm alongside a dynamic scheduling strategy to handle the huge amounts of walking queries. FlowWalker stands as a memory-efficient framework that requires no auxiliary data structures in GPU global memory. We examine the performance of FlowWalker extensively on ten datasets, and experiment results show that FlowWalker achieves up to 752.2x, 72.1x, and 16.4x speedup compared with existing CPU, GPU, and FPGA random walk frameworks, respectively. Case study shows that FlowWalker diminishes random walk time from 35% to 3% in a pipeline of ByteDance friend recommendation GNN training.

cs.DC↗

Behind the Veil: Enhanced Indoor 3D Scene Reconstruction with Occluded Surfaces Completion

In this paper, we present a novel indoor 3D reconstruction method with occluded surface completion, given a sequence of depth readings. Prior state-of-the-art (SOTA) methods only focus on the reconstruction of the visible areas in a scene, neglecting the invisible areas due to the occlusions, e.g., the contact surface between furniture, occluded wall and floor. Our method tackles the task of completing the occluded scene surfaces, resulting in a complete 3D scene mesh. The core idea of our method is learning 3D geometry prior from various complete scenes to infer the occluded geometry of an unseen scene from solely depth measurements. We design a coarse-fine hierarchical octree representation coupled with a dual-decoder architecture, i.e., Geo-decoder and 3D Inpainter, which jointly reconstructs the complete 3D scene geometry. The Geo-decoder with detailed representation at fine levels is optimized online for each scene to reconstruct visible surfaces. The 3D Inpainter with abstract representation at coarse levels is trained offline using various scenes to complete occluded surfaces. As a result, while the Geo-decoder is specialized for an individual scene, the 3D Inpainter can be generally applied across different scenes. We evaluate the proposed method on the 3D Completed Room Scene (3D-CRS) and iTHOR datasets, significantly outperforming the SOTA methods by a gain of 16.8% and 24.2% in terms of the completeness of 3D reconstruction. 3D-CRS dataset including a complete 3D mesh of each scene is provided at project webpage.

cs.CV↗

Cosmological Fisher forecasts for next-generation spectroscopic surveys

Next-generation spectroscopic surveys such as the MegaMapper, MUltiplexed Survey Telescope (MUST), MaunaKea Spectroscopic Explorer (MSE), and Wide Spectroscopic Telescope (WST) are foreseen to increase the number of galaxy/quasar redshifts by an order of magnitude, with hundred millions of spectra that will be measured at $z>2$. We perform a Fisher matrix analysis for these surveys on the baryonic acoustic oscillation (BAO), the redshift-space distortion (RSD) measurement, the non-Gaussianity amplitude $f_{\rm NL}$, and the total neutrino mass $M_ν$. For BAO and RSD parameters, these surveys may achieve precision at sub-percent level (<0.5 per cent), representing an improvement of factor 10 w.r.t. the latest database. For NG, these surveys may reach an accuracy of $σ(f_{\rm NL})\sim 1$. They can also put a tight constraint on $M_ν$ with $σ(M_ν) \sim 0.02\,\rm eV$ if we do joint analysis with Planck and even $ 0.01\,\rm eV$ if combined with other data. In addition, we introduce a general survey model, to derive the cosmic volume and number density of tracers, given instrumental facilities and survey strategy. Using our Fisher formalism, we can explore (continuously) a wide range of survey observational parameters, and propose different survey strategies that optimise the cosmological constraints. Fixing the fibre number and survey duration, we show that the best strategy for $f_{\rm NL}$ and $M_ν$ measurement is to observe large volumes, despite the noise increase. However, the strategy differs for the apparent magnitude limit. Finally, we prove that increasing the fibre number improves $M_ν$ measurement but not significantly $f_{\rm NL}$.

astro-ph.CO↗