arXiv subjects

Chen-Yu Wang

Publications and source records attributed to Chen-Yu Wang.

At least 19 recordsLinked to original sources

arXiv2026

VeCAS: Vessel-Focused Contrast-Free Angiogram Synthesis for Vascular Interventions

X-ray angiography relies on iodinated contrast agents to visualize vascular structures during image-guided interventions. However, contrast administration carries risks of adverse events, motivating the development of contrast-free alternatives. Generating X-ray angiograms directly from non-contrast X-ray images offers a potential solution, but existing approaches remain limited by (i) insufficient control over vascular localization and (ii) inefficient modeling of redundant background content. To address these challenges, we propose VeCAS, a two-stage vessel-focused contrast-free angiogram synthesis framework that separates vascular structure localization from angiographic appearance synthesis. In Stage I, a discriminative model localizes vascular structures in non-contrast X-ray images, while cross-modality latent distillation transfers vessel-sensitive knowledge from X-ray angiograms during training. In Stage II, a vessel-focused inpainting model synthesizes angiographic appearance within the localized vascular regions while preserving the non-vascular background. Experiments on an in-house lower-limb vascular intervention dataset show that VeCAS outperforms the comparison methods in terms of vascular structural fidelity and image quality. Visual Turing tests and physician assessments indicate the perceptual realism of the synthesized angiograms. In addition, robotic guidewire navigation experiments in vascular phantoms show that VeCAS guidance reduces the time to target by 41.4% and the number of operation steps by 40.7% compared with non-contrast guidance. Together, these results suggest the potential of VeCAS to serve as ``meta contrast agent'' for vascular interventions.

cs.CV↗

arXiv2026

On Explicit Super-Expressive Approximation for Neural Networks

In this work, we investigate the fixed-architecture neural network approximation with explicit parameter bounds and elementary activations. While prior work demonstrated super-expressive approximation using fixed-size networks, they lack quantitative and non-asymptotic characterizations of parameter magnitude with respect to the approximation error. We resolve this issue by introducing the Chinese Remainder Theorem as a constructive encoding mechanism. For Lipschitz continuous functions on $[0,1]^D$, we construct a width-$\max\{D,4\}$, depth-$5$ network with explicit parameter-error trade-offs. For H\"older-smooth functions in $C^{r,\gamma}_A\left([0,1]^D\right)$, our fixed network of width $\max\{2D,\ D+5N+1\}$ and depth $r + 9$ achieves the parameter magnitude $\mathcal{P}$ bounded by $\log_2 \mathcal{P}=\mathcal{O}\bigl(\varepsilon^{-2D/(r+\gamma)}\log(1/\varepsilon)\bigr)$. This is the dual result compared to those in the parameter-bounded and architecture-unbounded paradigm.

cs.LG↗

arXiv2026

Amortized Inverse Kinematics via Graph Attention for Real-Time Human Avatar Animation

Inverse kinematics (IK) is a core operation in animation, robotics, and biomechanics: given Cartesian constraints, recover joint rotations under a known kinematic tree. In many real-time human avatar pipelines, the available signal per frame is a sparse set of tracked 3D joint positions, whereas animation systems require joint orientations to drive skinning. Recovering full orientations from positions is underconstrained, most notably because twist about bone axes is ambiguous, and classical IK solvers typically rely on iterative optimization that can be slow and sensitive to noisy inputs. We introduce IK-GAT, a lightweight graph-attention network that reconstructs full-body joint orientations from 3D joint positions in a single forward pass. The model performs message passing over the skeletal parent-child graph to exploit kinematic structure during rotation inference. To simplify learning, IK-GAT predicts rotations in a bone-aligned world-frame representation anchored to rest-pose bone frames. This parameterization makes the twist axis explicit and is exactly invertible to standard parent-relative local rotations given the kinematic tree and rest pose. The network uses a continuous 6D rotation representation and is trained with a geodesic loss on SO(3) together with an optional forward-kinematics consistency regularizer. IK-GAT produces animation-ready local rotations that can directly drive a rigged avatar or be converted to pose parameters of SMPL-like body models for real-time and online applications. With 374K parameters and over 650 FPS on CPU, IK-GAT outperforms VPoser-based per-frame iterative optimization without warm-start at significantly lower cost, and is robust to initial pose and input noise

cs.CV↗

arXiv2026

Probing the Reliability of Driving VLMs: From Inconsistent Responses to Grounded Temporal Reasoning

A reliable driving assistant should provide consistent responses based on temporally grounded reasoning derived from observed information. In this work, we investigate whether Vision-Language Models (VLMs), when applied as driving assistants, can response consistantly and understand how present observations shape future outcomes, or whether their outputs merely reflect patterns memorized during training without temporally grounded reasoning. While recent efforts have integrated VLMs into autonomous driving, prior studies typically emphasize scene understanding and instruction generation, implicitly assuming that strong visual interpretation naturally enables consistant future reasoning and thus ensures reliable decision-making, a claim we critically examine. We focus on two major challenges limiting VLM reliability in this setting: response inconsistency, where minor input perturbations yield different answers or, in some cases, responses degenerate toward near-random guessing, and limited temporal reasoning, in which models fail to reason and align sequential events from current observations, often resulting in incorrect or even contradictory responses. Moreover, we find that models with strong visual understanding do not necessarily perform best on tasks requiring temporal reasoning, indicating a tendency to over-rely on pretrained patterns rather than modeling temporal dynamics. To address these issues, we adopt existing evaluation methods and introduce FutureVQA, a human-annotated benchmark dataset specifically designed to assess future scene reasoning. In addition, we propose a simple yet effective self-supervised tuning approach with chain-of-thought reasoning that improves both consistency and temporal reasoning without requiring temporal labels.

cs.CV↗

arXiv2026

Vascular anatomy-aware self-supervised pre-training for X-ray angiogram analysis

X-ray angiography is the gold standard imaging modality for cardiovascular diseases. However, current deep learning approaches for X-ray angiogram analysis are severely constrained by the scarcity of annotated data. While large-scale self-supervised learning (SSL) has emerged as a promising solution, its potential in this domain remains largely unexplored, primarily due to the lack of effective SSL frameworks and large-scale datasets. To bridge this gap, we introduce a vascular anatomy-aware masked image modeling (VasoMIM) framework that explicitly integrates domain-specific anatomical knowledge. Specifically, VasoMIM comprises two key designs: an anatomy-guided masking strategy and an anatomical consistency loss. The former strategically masks vessel-containing patches to compel the model to learn robust vascular semantics, while the latter preserves structural consistency of vessels between original and reconstructed images, enhancing the discriminability of the learned representations. In conjunction with VasoMIM, we curate XA-170K, the largest X-ray angiogram pre-training dataset to date. We validate VasoMIM on four downstream tasks across six datasets, where it demonstrates superior transferability and achieves state-of-the-art performance compared to existing methods. These findings highlight the significant potential of VasoMIM as a foundation model for advancing a wide range of X-ray angiogram analysis tasks. VasoMIM and XA-170K will be available at https://github.com/Dxhuang-CASIA/XA-SSL.

cs.CV↗

arXiv2025

OverLayBench: A Benchmark for Layout-to-Image Generation with Dense Overlaps

Despite steady progress in layout-to-image generation, current methods still struggle with layouts containing significant overlap between bounding boxes. We identify two primary challenges: (1) large overlapping regions and (2) overlapping instances with minimal semantic distinction. Through both qualitative examples and quantitative analysis, we demonstrate how these factors degrade generation quality. To systematically assess this issue, we introduce OverLayScore, a novel metric that quantifies the complexity of overlapping bounding boxes. Our analysis reveals that existing benchmarks are biased toward simpler cases with low OverLayScore values, limiting their effectiveness in evaluating model performance under more challenging conditions. To bridge this gap, we present OverLayBench, a new benchmark featuring high-quality annotations and a balanced distribution across different levels of OverLayScore. As an initial step toward improving performance on complex overlaps, we also propose CreatiLayout-AM, a model fine-tuned on a curated amodal mask dataset. Together, our contributions lay the groundwork for more robust layout-to-image generation under realistic and challenging scenarios. Project link: https://mlpc-ucsd.github.io/OverLayBench.

cs.CV↗

arXiv2025

Seeing Clearly, Forgetting Deeply: Revisiting Fine-Tuned Video Generators for Driving Simulation

Recent advancements in video generation have substantially improved visual quality and temporal coherence, making these models increasingly appealing for applications such as autonomous driving, particularly in the context of driving simulation and so-called "world models". In this work, we investigate the effects of existing fine-tuning video generation approaches on structured driving datasets and uncover a potential trade-off: although visual fidelity improves, spatial accuracy in modeling dynamic elements may degrade. We attribute this degradation to a shift in the alignment between visual quality and dynamic understanding objectives. In datasets with diverse scene structures within temporal space, where objects or perspective shift in varied ways, these objectives tend to highly correlated. However, the very regular and repetitive nature of driving scenes allows visual quality to improve by modeling dominant scene motion patterns, without necessarily preserving fine-grained dynamic behavior. As a result, fine-tuning encourages the model to prioritize surface-level realism over dynamic accuracy. To further examine this phenomenon, we show that simple continual learning strategies, such as replay from diverse domains, can offer a balanced alternative by preserving spatial accuracy while maintaining strong visual quality.

cs.CV↗

arXiv2024

Triple real-emission contribution to the zero-jettiness soft function at N3LO in QCD

Recently, we have presented the result for the zero-jettiness soft function at next-to-next-to-next-to-leading order (N3LO) in perturbative QCD [arXiv:2409.11042], without providing technical details of the calculation. The goal of this paper is to describe the most important element of that computation, the triple real-emission contribution. We present a detailed discussion of the many technical aspects of the calculation, for which a number of methodological innovations was required. Although some elements of the calculation were discussed earlier [arXiv:2004.03285,arXiv:2206.12323,arXiv:2111.13594,arXiv:2204.09459,arXiv:2401.05245], this paper is intended to provide a complete summary of the methods used in the computation of the triple real-emission contribution to the soft function.

hep-ph↗

arXiv2024

Bridging the Gap: GRB 230812B -- A Three-Second Supernova-Associated Burst Detected by the GRID Mission

GRB 230812B, detected by the Gamma-Ray Integrated Detectors (GRID) constellation mission, is an exceptionally bright gamma-ray burst (GRB) with a duration of only 3 seconds. Sitting near the traditional boundary ($\sim$ 2 s) between long and short GRBs, GRB 230812B is notably associated with a supernova (SN), indicating a massive star progenitor. This makes it a rare example of a short-duration GRB resulting from stellar collapse. Our analysis, using a time-evolving synchrotron model, suggests that the burst has an emission radius of approximately $10^{14.5}$~cm. We propose that the short duration of GRB 230812B is due to the combined effects of the central engine's activity time and the time required for the jet to break through the stellar envelope. Our findings provide another case that challenges the conventional view that short-duration GRBs originate exclusively from compact object mergers, demonstrating that a broader range of durations exists for GRBs arising from the collapse of massive stars.

astro-ph.HE↗

arXiv2024

Zero-jettiness soft function to third order in perturbative QCD

We present the high-precision result for the zero-jettiness soft function at next-to-next-to-next-to-leading order (N3LO) in perturbative QCD. At this perturbative order, the soft function is the last missing ingredient required for the computation of a hadronic colour singlet production or a colour singlet decay into two jets using the zero-jettiness variable as the slicing parameter. Furthermore, the knowledge of the N3LO soft function enables the re-summed description of the thrust distribution in the process $e^+ e^- \to \textrm{hadrons}$ through next-to-next-to-next-to-leading logarithmic order, which is important for the extraction of the strong coupling constant using this shape variable. On the methodological side, the complexity of the zero-jettiness variable forced us to develop a new semi-analytic method for phase-space integration in the presence of constraints parameterized through Heaviside functions which, hopefully, will be useful for further development of the $N$-jettiness slicing scheme.

hep-ph↗

arXiv2024

$Z_c$ and $Z_{cs}$ systems with operator mixing at NLO in QCD sum rules

We study the mass spectra of hidden-charm tetraquark systems with quantum numbers $(I^G)J^P=(1^+)1^+$ using QCD sum rules. The analysis incorporates the complete next-to-leading order (NLO) contribution to the perturbative QCD part of the operator product expansions, with particular attention to operator mixing effects due to renormalization group evolution. For the $\bar{d}c\bar{c}u$ system, the masses of two mixed operators, $J_{1,5}^{\text{Mixed}}$ and $J_{2,6}^{\text{Mixed}}$, are determined to be $3.89^{+0.18}_{-0.12}$ GeV and $4.03^{+0.06}_{-0.07}$ GeV, respectively, closely matching those of $Z_c$(3900) and $Z_c(4020)$. Similarly, for the $\bar{s}c\bar{c}u$ states, the masses of $J_{1,5}^{\text{Mixed}}$ and $J_{2,6}^{\text{Mixed}}$ are found to be $4.02^{+0.17}_{-0.09}$ GeV and $4.21^{+0.08}_{-0.07}$ GeV, respectively, in close proximity to $Z_{cs}$(3983)/$Z_{cs}$(4000) and $Z_{cs}$(4220), consistent with the expectation that they are the partners of $Z_c$(3900) and $Z_c$(4020). Our results highlight the crucial role of operator mixing, an inevitable effect in a complete NLO calculation, in achieving a robust phenomenological description for the tetraquark system.

hep-ph↗

arXiv2024

One-loop corrections to the double-real emission contribution to the zero-jettiness soft function at N3LO in QCD

We present an analytic calculation of the one-loop correction to the double-real emission contribution to the zero-jettiness soft function at N3LO in QCD, accounting for both gluon-gluon and quark-antiquark soft final-state partons. We explain all the relevant steps of the computation including the reduction of phase-space integrals to master integrals in the presence of Heaviside functions, and the methods we employed to compute them.

hep-ph↗

arXiv2024

Reclassifying Feynman integrals as special functions

Although Feynman integrals in general cannot be expressed as well-studied special functions, they can be calculated systematically and efficiently using the \texttt{AMFlow} method in combination with differential equations in the kinematic space. Therefore, it is constructive to define Feynman integrals as a new class of special functions (or transcedental numbers if there is no kinematic variable). Several crucial avenues for further exploration in this direction are outlined.

hep-ph↗

arXiv2023

Homoclinic orbits in Kerr-Newman black holes

We present the exact solutions of the homoclinic orbits for the timelike geodesics of the particle on the general nonequatorial orbits in the Kerr-Newman black holes. The homoclinic orbit is the separatrix between bound and plunging geodesics, a solution that asymptotes to an energetically bound, unstable spherical orbit. The solutions are written in terms of the elliptical integrals and the Jacobi elliptic functions of manifestly real functions of the Mino time where we focus on the effect from the charge of the black hole to the homoclinic orbits. The parameter space of the homoclinic solutions is explored. The nonequatorial homoclinic orbits in Kerr cases can be obtained by setting the charge of the black holes to be zero. The homoclinic orbits and the associated phase portrait as a function of the radial position and its derivation with respect to the Mino time are plotted using the analytical solutions. In particular, the solutions can reduce to the zero azimuthal angular moment homoclinic orbits for understanding the frame dragging effects from the spin as well as the charge of the black hole. The implications of the obtained results to observations are discussed.

gr-qc↗

arXiv2023

GRB 221009A: An ordinary nearby GRB with extraordinary observational properties

The gamma-ray burst GRB 221009A, known as the ``brightest-of-all-time" (BOAT), is the closest energetic burst detected so far, with an energy of $E_{γ,\rm iso} \sim 10^{55}$ ergs. This study aims to assess its compatibility with known GRB energy and luminosity distributions. Our analysis indicates that the energy/luminosity function of GRBs is consistent across various redshift intervals, and that the inclusion of GRB 221009A does not significantly impact the function at low redshifts. Additionally, our evaluation of the best-fitting result of the entire GRB sample suggests that the expected number of GRBs with energy greater than $10^{55}$ ergs at a low redshift is 0.2, so that the emergence of GRB 221009A is consistent with expected energy/luminosity functions within $\sim 2σ$ Poisson fluctuation error, still adhering to the principles of small number statistics. Furthermore, we find that GRB 221009A and other energetic bursts, defined as $E_{γ,\rm iso} \gtrsim10^{54} {\rm ergs}$, exhibit no significant differences in terms of distributions of $T_{90}$, minimum timescale, Amati relation, $E_{\rm γ,iso}$-$E_{\rm X,iso}$ relation, $L_{γ,\rm iso}-Γ_0$ relation, $E_{γ,\rm iso}-Γ_0$ relation, $L_{γ,\rm iso}-E_{\rm p,i}-Γ_0$ relation, and host galaxy properties, compared to normal long GRBs. This suggests that energetic GRBs (including GRB 221009A) and other long GRBs likely have similar progenitor systems and undergo similar energy dissipation and radiation processes. The generation of energetic GRBs may be due to more extreme central engine properties or, more likely, a rarer viewing configuration of a quasi-universal structured jet.

astro-ph.HE↗

arXiv2022

NLO results with operator mixing for fully heavy tetraquarks in QCD sum rules

We study the mass spectra of $\bar{Q}Q\bar{Q}Q\ (Q=c,b)$ systems in QCD sum rules with the complete next-to-leading order (NLO) contribution to the perturbative QCD part of the correlation functions. Instead of meson-meson or diquark-antidiquark currents, we use diagonalized currents under operator renormalization. We find that differing from conventional mesons $\bar qq$ and baryons $qqq$, a unique feature of the multiquark systems like $\bar{Q}Q\bar{Q}Q$ is the operator mixing or color configuration mixing induced by NLO corrections, which is crucial to understand the color structure of the states. Our numerical results show that the NLO corrections are very important for the $\bar{Q}Q\bar{Q}Q$ system, because they not only give significant contributions but also reduce the scheme and scale dependence and make Borel platform more distinct, especially for the $\bar{b}b\bar{b}b$ in the $\overline{\rm{MS}}$ scheme. We use currents that have good perturbation convergence in our phenomenological analysis. With the $\overline{\rm{MS}}$ scheme, we get three $J^{PC}=0^{++}$ states, with masses $6.35^{+0.20}_{-0.17}$ GeV, $6.56^{+0.18}_{-0.20}$ GeV and $6.95^{+0.21}_{-0.31}$ GeV, respectively. The first two seem to agree with the broad structure around $6.2\sim6.8$ GeV measured by the LHCb collaboration in the $J/ψJ/ψ$ spectrum, and the third seems to agree with the narrow resonance $X(6900)$. For the $2^{++}$ states we find one with mass $7.03^{+0.22}_{-0.26}$ GeV, which is also close to that of $X(6900)$, and another one around $7.25^{+0.21}_{-0.35}$ GeV, which has good scale dependence but slightly large scheme dependence.

hep-ph↗

arXiv2022

Null and time-like geodesics in Kerr-Newman black hole exterior

We study the null and time-like geodesics of the light and the neutral particles respectively in the exterior of Kerr-Newman black holes. The geodesic equations are known to be written as a set of first-order differential equations in Mino time from which the angular and radial potentials can be defined. We classify the roots for both potentials, and mainly focus on those of the radial potential with an emphasis on the effect from the charge of the black holes. We then obtain the solutions of the trajectories in terms of the elliptical integrals and the Jacobian elliptic functions for both null and time-like geodesics, which are manifestly real functions of the Mino time that the initial conditions can be explicitly specified. We also describe the details of how to reduce those solutions into the cases of the spherical orbits. The effect of the black hole's charge decreases the radii of the spherical motion of the light and the particle for both direct and retrograde motions. In particular, we focus on the light/particle boomerang of the spherical orbits due to the frame dragging from the back hole's spin with the effect from the charge of the black hole. To sustain the change of the azimuthal angle of the light rays, say for example $Δϕ=π$ during the whole trip, the presence of the black hole's charge decreases the radius of the orbit and consequently reduces the needed values of the black hole's spin. As for the particle boomerang, the particle's inertia renders smaller change of the angle $Δϕ$ as compared with the light boomerang. Moreover, the black hole's charge also results in the smaller angle change $Δϕ$ of the particle than that in the Kerr case. The implications of the obtained results to observations are discussed.

gr-qc↗

arXiv2022

Non-factorisable contributions to $t$-channel single-top production at the LHC and FCC

Single top quark is mainly produced through the $t$-channel W boson exchange $q + b \rightarrow q' + t$ at LHC. This process probes Wtb vertex directly and can be used to measure the CKM matrix element $V_{tb}$ or to constrain the bottom quark PDF. The non-factorisable contributions are the last missing piece of the NNLO corrections. In these proceedings, we discuss in a first part the ostensible importance of such corrections and the obtention of the different relevant amplitudes. In the second part, recently published results at the energy of the LHC are compared to new results for proton-proton collision at $100 \: {\rm TeV}$, the energy of the FCC.

hep-ph↗