arXiv subjects

Yuchen Du

Publications and source records attributed to Yuchen Du.

At least 19 recordsLinked to original sources

arXiv2026

Reviewing the Reviewer: Graph-Enhanced LLMs for E-commerce Appeal Adjudication

Hierarchical review workflows, where a second-tier reviewer (Checker) corrects first-tier (Maker) decisions, generate valuable correction signals that encode why initial judgments failed. However, learning from these signals is hindered by information asymmetry: corrections often depend on verification actions unavailable to Makers or automated systems. We address this challenge by introducing explicit action modeling as an inferential constraint that grounds reasoning in verifiable operations rather than unconstrained text generation. We propose the Evidence-Action-Factor-Decision (EAFD) schema, a minimal representation for adjudication reasoning that prevents hallucination through operational grounding and enables learning from correction signals via explicit conflict modeling. Building on this schema, we develop a conflict-aware graph reasoning framework that: (1) constructs EAFD graphs from historical cases capturing Maker-Checker disagreements, (2) aggregates them into a retrievable knowledge base, and (3) performs top-down deductive reasoning for new cases by projecting validated resolution paths from precedents. A distinctive capability is the Request More Information (RMI) outcome: when evidence is insufficient, the system identifies precisely which verification actions remain unexecuted and generates targeted information requests. We evaluate the framework in large-scale e-commerce seller appeal adjudication. While a standard LLM-only baseline achieves only 70.8% alignment with human experts, incorporating action modeling with RMI improves alignment to 87.5%. Augmenting this with the retrieval-based knowledge graph yields the best offline performance of 95.8%. Following online deployment, the framework maintains robust performance, achieving a 96.3% alignment rate in production, demonstrating its real-world effectiveness.

cs.CL↗

arXiv2026

Conservative binary dynamics to third post-Minkowskian order beyond General Relativity

We present the conservative dynamics of compact binaries to third order in the post-Minkowskian approximation in a theory that extends general relativity by a massless scalar field coupled to the Gauss-Bonnet invariant. We employ the effective field theory approach to construct the effective action of binary systems by integrating out the metric and scalar degrees of freedom that mediate the gravitational interactions between the two bodies. We derive analytical expressions for the scattering impulse and the deflection angle to third order in the post-Minkowskian expansion. Our results are found to be in agreement, in the overlapping regimes, with state-of-the-art calculations in the post-Newtonian/post-Minkowskian theory.

gr-qc↗

arXiv2025

Signal-SGN++: Topology-Enhanced Time-Frequency Spiking Graph Network for Skeleton-Based Action Recognition

Graph Convolutional Networks (GCNs) demonstrate strong capability in modeling skeletal topology for action recognition, yet their dense floating-point computations incur high energy costs. Spiking Neural Networks (SNNs), characterized by event-driven and sparse activation, offer energy efficiency but remain limited in capturing coupled temporal-frequency and topological dependencies of human motion. To bridge this gap, this article proposes Signal-SGN++, a topology-aware spiking graph framework that integrates structural adaptivity with time-frequency spiking dynamics. The network employs a backbone composed of 1D Spiking Graph Convolution (1D-SGC) and Frequency Spiking Convolution (FSC) for joint spatiotemporal and spectral feature extraction. Within this backbone, a Topology-Shift Self-Attention (TSSA) mechanism is embedded to adaptively route attention across learned skeletal topologies, enhancing graph-level sensitivity without increasing computational complexity. Moreover, an auxiliary Multi-Scale Wavelet Transform Fusion (MWTF) branch decomposes spiking features into multi-resolution temporal-frequency representations, wherein a Topology-Aware Time-Frequency Fusion (TATF) unit incorporates structural priors to preserve topology-consistent spectral fusion. Comprehensive experiments on large-scale benchmarks validate that Signal-SGN++ achieves superior accuracy-efficiency trade-offs, outperforming existing SNN-based methods and achieving competitive results against state-of-the-art GCNs under substantially reduced energy consumption.

cs.CV↗

arXiv2025

SNN-Driven Multimodal Human Action Recognition via Sparse Spatial-Temporal Data Fusion

Multimodal human action recognition based on RGB and skeleton data fusion, while effective, is constrained by significant limitations such as high computational complexity, excessive memory consumption, and substantial energy demands, particularly when implemented with Artificial Neural Networks (ANN). These limitations restrict its applicability in resource-constrained scenarios. To address these challenges, we propose a novel Spiking Neural Network (SNN)-driven framework for multimodal human action recognition, utilizing event camera and skeleton data. Our framework is centered on two key innovations: (1) a novel multimodal SNN architecture that employs distinct backbone networks for each modality-an SNN-based Mamba for event camera data and a Spiking Graph Convolutional Network (SGN) for skeleton data-combined with a spiking semantic extraction module to capture deep semantic representations; and (2) a pioneering SNN-based discretized information bottleneck mechanism for modality fusion, which effectively balances the preservation of modality-specific semantics with efficient information compression. To validate our approach, we propose a novel method for constructing a multimodal dataset that integrates event camera and skeleton data, enabling comprehensive evaluation. Extensive experiments demonstrate that our method achieves superior performance in both recognition accuracy and energy efficiency, offering a promising solution for practical applications.

cs.CV↗

arXiv2024

Worldline Formalism, Eikonal Expansion and the Classical Limit of Scattering Amplitudes

We revisit the fundamentals of two different methods for calculating classical observables: the eikonal method, which is a scattering amplitude-based method, and the worldline quantum field theory (WQFT) method. The latter has been considered an extension of the worldline effective field theory. We show that the eikonal and WQFT methods are equivalent and that calculations can be translated freely between them. Concretely, we focus on 2-into-2 scattering processes mediated by massless force carriers. On the one hand, taking the classical limit of the QFT scattering amplitude leads to the eikonal method. On the other hand, since in the classical limit the scattering particles are almost on-shell throughout the scattering process, the worldline, a first quantized formalism, is the most efficient framework to study the scattering amplitude. This is an alternate but equivalent formalism to the quantum field theoretic (QFT) framework. By taking the classical limit of the scattering amplitude computed in the worldline, we can derive the WQFT rules of Mogull, Plefka and Steinhoff. In WQFT, the Feynman diagrams are reorganized into a new set of diagrams that facilitate the $\hbar$ expansion. Unlike the QFT eikonal method, which works recursively in identifying the eikonal phase, the worldline-based computation allows to target and systematically extract the classical contributions directly through a specific set of WQFT diagrams. In worldline formalism the perturbative expansion of the scattering amplitude is naturally organized in diagrams which factorize (reducible) and diagrams which are new to that order (irreducible), in a one-to-one map with the structure of the amplitude in the eikonal method. This opened up the possibility to investigate and prove the conjectured exponentiation of the eikonal phase in arXiv: 2409.12895.

hep-th↗

arXiv2024

Worldline Proof of Eikonal Exponentiation

In this paper, working in the eikonal approximation, we present a proof for the exponentiation of the 2-body eikonal phase to {\it all orders in the eikonal expansion}, for scalar particles interacting electromagnetically or gravitationally. The proof is based on the worldline formalism, which is an alternative, first quantized method to the standard QFT calculation of the scattering amplitude. We show that in the worldline formalism the 2-body scattering amplitude written in impact parameter space naturally factorizes at each loop order. This factorization is responsible for the exponentiation of the eikonal phase, a result which was anticipated in the work of Mogull, Plefka, and Steinhoff [2010.02865 [hep-th]].

hep-th↗

arXiv2024

Signal-SGN: A Spiking Graph Convolutional Network for Skeletal Action Recognition via Learning Temporal-Frequency Dynamics

For multimodal skeleton-based action recognition, Graph Convolutional Networks (GCNs) are effective models. Still, their reliance on floating-point computations leads to high energy consumption, limiting their applicability in battery-powered devices. While energy-efficient, Spiking Neural Networks (SNNs) struggle to model skeleton dynamics, leading to suboptimal solutions. We propose Signal-SGN (Spiking Graph Convolutional Network), which utilizes the temporal dimension of skeleton sequences as the spike time steps and represents features as multi-dimensional discrete stochastic signals for temporal-frequency domain feature extraction. It combines the 1D Spiking Graph Convolution (1D-SGC) module and the Frequency Spiking Convolution (FSC) module to extract features from the skeleton represented as spiking form. Additionally, the Multi-Scale Wavelet Transform Feature Fusion (MWTF) module is proposed to extract dynamic spiking features and capture frequency-specific characteristics, enhancing classification performance. Experiments across three large-scale datasets reveal Signal-SGN exceeding state-of-the-art SNN-based methods in accuracy and computational efficiency while attaining comparable performance with GCN methods and significantly reducing theoretical energy consumption.

cs.CV↗

arXiv2024

MK-SGN: A Spiking Graph Convolutional Network with Multimodal Fusion and Knowledge Distillation for Skeleton-based Action Recognition

In recent years, multimodal Graph Convolutional Networks (GCNs) have achieved remarkable performance in skeleton-based action recognition. The reliance on high-energy-consuming continuous floating-point operations inherent in GCN-based methods poses significant challenges for deployment in energy-constrained, battery-powered edge devices. To address these limitations, MK-SGN, a Spiking Graph Convolutional Network with Multimodal Fusion and Knowledge Distillation, is proposed to leverage the energy efficiency of Spiking Neural Networks (SNNs) for skeleton-based action recognition for the first time. By integrating the energy-saving properties of SNNs with the graph representation capabilities of GCNs, MK-SGN achieves significant reductions in energy consumption while maintaining competitive recognition accuracy. Firstly, we formulate a Spiking Multimodal Fusion (SMF) module to effectively fuse multimodal skeleton data represented as spike-form features. Secondly, we propose the Self-Attention Spiking Graph Convolution (SA-SGC) module and the Spiking Temporal Convolution (STC) module, to capture spatial relationships and temporal dynamics of spike-form features. Finally, we propose an integrated knowledge distillation strategy to transfer information from the multimodal GCN to the SGN, incorporating both intermediate-layer distillation and soft-label distillation to enhance the performance of the SGN. MK-SGN exhibits substantial advantages, surpassing state-of-the-art GCN frameworks in energy efficiency and outperforming state-of-the-art SNN frameworks in recognition accuracy. The proposed method achieves a remarkable reduction in energy consumption, exceeding 98\% compared to conventional GCN-based approaches. This research establishes a robust baseline for developing high-performance, energy-efficient SNN-based models for skeleton-based action recognition

cs.CV↗

arXiv2024

Understanding Social Perception, Interactions, and Safety Aspects of Sidewalk Delivery Robots Using Sentiment Analysis

This article presents a comprehensive sentiment analysis (SA) of comments on YouTube videos related to Sidewalk Delivery Robots (SDRs). We manually annotated the collected YouTube comments with three sentiment labels: negative (0), positive (1), and neutral (2). We then constructed models for text sentiment classification and tested the models' performance on both binary and ternary classification tasks in terms of accuracy, precision, recall, and F1 score. Our results indicate that, in binary classification tasks, the Support Vector Machine (SVM) model using Term Frequency-Inverse Document Frequency (TF-IDF) and N-gram get the highest accuracy. In ternary classification tasks, the model using Bidirectional Encoder Representations from Transformers (BERT), Long Short-Term Memory Networks (LSTM) and Gated Recurrent Unit (GRU) significantly outperforms other machine learning models, achieving an accuracy, precision, recall, and F1 score of 0.78. Additionally, we employ the Latent Dirichlet Allocation model to generate 10 topics from the comments to explore the public's underlying views on SDRs. Drawing from these findings, we propose targeted recommendations for shaping future policies concerning SDRs. This work provides valuable insights for stakeholders in the SDR sector regarding social perception, interaction, and safety.

cs.RO↗

arXiv2023

Gravitational wave energy-momentum tensor and radiated power in a strongly curved background

Allowing for the possibility of extra dimensions, there are two paradigms: either the extra dimensions are hidden from observations by being compact and small as in Kaluza-Klein scenarios, or the extra dimensions are large/non-compact and undetectable due to a large warping as in the Randall-Sundrum scenario. In the latter case, the five-dimensional background has a large curvature, and Isaacson's construction of the gravitational energy-momentum tensor, which relies on the assumption that the wavelength of the metric fluctuations is much smaller than the curvature length of the background spacetime, cannot be used. In this paper, we construct the gravitational energy-momentum tensor in a strongly curved background such as Randall-Sundrum. We perform a scalar-vector-tensor decomposition of the metric fluctuations with respect to the $SO(1,3)$ background isometry and construct the covariantly-conserved gravitational energy-momentum tensor out of the gauge-invariant metric fluctuations. We give a formula for the power radiated by gravitational waves and verify it in known cases. In using the gauge-invariant metric fluctuations to construct the gravitational energy-momentum tensor we follow previous work done in cosmology. Our framework has applicability beyond the Randall-Sundrum model.

gr-qc↗

arXiv2023

Analytical model for large-scale design of sidewalk delivery robot systems

With the rise in demand for local deliveries and e-commerce, robotic deliveries are being considered as efficient and sustainable solutions. However, the deployment of such systems can be highly complex due to numerous factors involving stochastic demand, stochastic charging and maintenance needs, complex routing, etc. We propose a model that uses continuous approximation methods for evaluating service trade-offs that consider the unique characteristics of large-scale sidewalk delivery robot systems used to serve online food deliveries. The model captures both the initial cost and the operation cost of the delivery system and evaluates the impact of constraints and operation strategies on the deployment. By minimizing the system cost, variables related to the system design can be determined. First, the minimization problem is formulated based on a homogeneous area, and the optimal system cost can be derived as a closed-form expression. By evaluating the expression, relationships between variables and the system cost can be directly obtained. We then apply the model in neighborhoods in New York City to evaluate the cost of deploying the sidewalk delivery robot system in a real-world scenario. The results shed light on the potential of deploying such a system in the future.

cs.CY↗

arXiv2023

Tree-level Graviton Scattering in the Worldline Formalism

We use the worldline formalism to study tree-level scattering processes involving gravitons. A massless spin 2 particle is described by an $N=4$ supersymmetric worldline action which is also $O(4)$ symmetric. More generally, $N=2S$ supersymmetric worldline actions exhibiting $O(N)$ symmetry describe free spin $S$ particles. Recently a BRST approach was used to construct the on-shell background graviton emission vertex from a graviton worldline. Nonetheless, an action describing the coupling of higher spin ($S\geq 2$) particles with generic background gravity is unknown. In this paper, we found that in order to reproduce Einstein's general relativity 3-point graviton vertex, interpreted as the emission of an off-shell graviton from the worldline, the coupling to background gravity must break the $O(4)$ symmetry to $O(2) \times O(2)$. In addition to this symmetry-breaking feature, we also found that the coefficient $β$ of the worldline action counterterm $βR$ differs from previous results in the literature. By comparing the linearized graviton and photon emission vertex operators from different worldlines, we noticed that they obey a squaring relation. For MHV (Maximal Helicity Violating) amplitudes, these squaring relations among the linearized vertex operators directly result in double-copy-like relations between the scattering amplitudes.

hep-th↗

arXiv2020

Probing Compactified Extra Dimensions with Gravitational Waves

We study the effect of compact extra dimensions on the gravitational wave luminosity and waveform. We consider a toy model, with a compactified fifth dimension, and matter confined on a brane. We work in the context of five dimensional ($5d$) general relativity, though we do make connections with the corresponding Kaluza-Klein effective $4d$ theory. We show that the luminosity of gravitational waves emitted in $5d$ gravity by a binary with the same characteristics (same masses and separation distance) as a $4d$ binary is 20.8\% less relative to the $4d$ case, to leading post-Newtonian order. The phase of the gravitational waveform differs by 26\% relative to the $4d$ case, to leading post-Newtonian order. Such a correction arises mainly due to the coupling between matter and dilaton field in the effective $4d$ picture and agrees with previous calculations when we set black holes' scalar charges to be those computed from the Kaluza-Klein reduction. The above corrections to the waveform and the luminosity are inconsistent with the gravitational-wave and binary pulsar observations and thus they effectively rule out the possibility of such a simple compactified higher dimensions scenario. We also comment on how our results change if there are several compactified extra dimensions, and show that the discrepancy with $4d$ general relativity only increases.

gr-qc↗

arXiv2018

Field-effect transistors made from solution-grown two-dimensional tellurene

The reliable production of two-dimensional crystals is essential for the development of new technologies based on 2D materials. However, current synthesis methods suffer from a variety of drawbacks, including limitations in crystal size and stability. Here, we report the fabrication of large-area, high-quality 2D tellurium (tellurene) using a substrate-free solution process. Our approach can create crystals with a process-tunable thickness, from monolayer to tens of nanometres, and with lateral sizes of up to 100 um. The chiral-chain van der Waals structure of tellurene gives rise to strong in-plane anisotropic properties and large thickness dependent shifts in Raman vibrational modes, which is not observed in other 2D layered materials. We also fabricate tellurene field-effect transistors, which exhibit air-stable performance at room temperature for over two months, on off ratios on the order of 106 and field-effect mobilities of around 700 cm2 per Vs. Furthermore, by scaling down the channel length and integrating with high-k dielectrics, transistors with a significant on-state current density of 1 A mm-1 are demonstrated.

cond-mat.mtrl-sci↗

arXiv2017

1D van der Waals Material Tellurium: Raman Spectroscopy under Strain and Magneto-transport

Experimental demonstrations of 1D van der Waals material tellurium have been presented by Raman spectroscopy under strain and magneto-transport. Raman spectroscopy measurements have been performed under strains along different principle axes. Pronounced strain response along c-axis is observed due to the strong intra-chain covalent bonds, while no strain response is obtained along a-axis due to the weak inter-chain van der Waals interaction. Magneto-transport results further verify its anisotropic property, resulting in dramatically distinct magneto-resistance behaviors in terms of three different magnetic field directions. Specifically, phase coherence length extracted from weak antilocalization effect, L$_Φ$ ~ T$^{-0.5}$, claims its 2D transport characteristics when an applied magnetic field is perpendicular to the thin film. In contrast, L$_Φ$ ~ T$^{-0.33}$ is obtained from universal conductance fluctuations once the magnetic field is along c-axis of Te, indicating its nature of 1D transport along the helical atomic chains. Our studies, which are obtained on high quality single crystal tellurium thin film, appear to serve as strong evidences of its 1D van der Waals structure from experimental perspectives. It is the aim of this paper to address this special concept that differs from the previous well-studied 1D nanowires or 2D van der Waals materials.

cond-mat.mtrl-sci↗

arXiv2016

Observation of Optical and Electrical In-plane Anisotropy in High-mobility Few-layer ZrTe5

Transition metal pentatelluride ZrTe5 is a versatile material in condensed-matter physics and has been intensively studied since the 1980s. The most fascinating feature of ZrTe5 is that it is a 3D Dirac semimetal which has linear energy dispersion in all three dimensions in momentum space. Structure-wise, ZrTe5 is a layered material held together by weak interlayer van der Waals force. The combination of its unique band structure and 2D atomic structure provides a fertile ground for more potential exotic physical phenomena in ZrTe5 related to 3D Dirac semimentals. However the physical properties of its few-layer form have yet to be thoroughly explored. Here we report strong optical and electrical in-plane anisotropy of mechanically exfoliated few-layer ZrTe5. Raman spectroscopy shows significant intensity change with sample orientations, and the behavior of angle-resolved phonon modes at the gamma point is explained by theoretical calculation. DC conductance measurement indicates a 50% of difference along different in-plane directions. The diminishing of resistivity anomaly in few-layer samples indicates the evolution of band structure with reduced thickness. Low-temperature Hall experiment sheds lights on more intrinsic anisotropic electrical transport, with hole mobility of 3,000 and 1,500 cm2/Vs along a-axis and c-axis respectively. Pronounced quantum oscillations in magneto-resistance are observed at low temperatures with highest electron mobility up to 44,000 cm2/Vs.

cond-mat.mtrl-sci↗

arXiv2016

Performance Enhancement of Black Phosphorus Field-Effect Transistors by Chemical Doping

In this letter, a new approach to chemically dope black phosphorus (BP) is presented, which significantly enhances the device performance of BP field-effect transistors for an initial period of 18 h, before degrading to previously reported levels. By applying 2,3,5,6-tetrafluoro-7,7,8,8-tetracyanoquinodimethane (F4-TCNQ), low ON-state resistance of 3.2 ohm.mm and high field-effect mobility of 229 cm2/Vs are achieved with a record high drain current of 532 mA/mm at a moderate channel length of 1.5 μm.

cond-mat.mtrl-sci↗

arXiv2016

Auxetic Black Phosphorus: A 2D Material with Negative Poisson's Ratio

The Poisson's ratio of a material characterizes its response to uniaxial strain. Materials normally possess a positive Poisson's ratio - they contract laterally when stretched, and expand laterally when compressed. A negative Poisson's ratio is theoretically permissible but has not, with few exceptions of man-made bulk structures, been experimentally observed in any natural materials. Here, we show that the negative Poisson's ratio exists in the low-dimensional natural material black phosphorus, and that our experimental observations are consistent with first principles simulations. Through application of uniaxial strain along zigzag and armchair directions, we find that both interlayer and intralayer negative Poisson's ratios can be obtained in black phosphorus. The phenomenon originates from the puckered structure of its in-plane lattice, together with coupled hinge-like bonding configurations.

cond-mat.mtrl-sci↗