arXiv subjects

Xi Chen

Publications and source records attributed to Xi Chen.

At least 451 records · Page 25Linked to original sources

arXiv2025

Digitized Counter-Diabatic Quantum Optimization for Bin Packing Problem

The bin packing problem, a classical NP-hard combinatorial optimization challenge, has emerged as a promising candidate for quantum computing applications. In this work, we address the one-dimensional bin packing problem (1dBPP) using a digitized counter-diabatic quantum algorithm (DC-QAOA), which incorporates counter-diabatic (CD) driving to reduce quantum resource requirements while maintaining high solution quality, outperforming traditional methods such as QAOA. We evaluate three ansatz schemes-DC-QAOA, a CD-inspired ansatz, and a CD-mixer ansatz-each integrating CD terms with distinct combinations of cost and mixer Hamiltonians, resulting in different DC-QAOA variants. Among these, the CD-mixer ansatz demonstrates superior performance, showing robustness across various iteration counts, layer depths, and Hamiltonian steps, while consistently producing the most accurate approximations to exact solutions. To validate our approach, we solve a 10-item 1dBPP instance on an IBM quantum computer, optimizing circuit structures through simulations. Despite constraints on circuit depth, the CD-mixer ansatz achieves high accuracy and a high likelihood of success. These findings establish DC-QAOA, particularly the CD-mixer variant, as a powerful framework for solving combinatorial optimization problems on near-term quantum devices.

quant-ph↗

arXiv2025

Step-Audio: Unified Understanding and Generation in Intelligent Speech Interaction

Real-time speech interaction, serving as a fundamental interface for human-machine collaboration, holds immense potential. However, current open-source models face limitations such as high costs in voice data collection, weakness in dynamic control, and limited intelligence. To address these challenges, this paper introduces Step-Audio, the first production-ready open-source solution. Key contributions include: 1) a 130B-parameter unified speech-text multi-modal model that achieves unified understanding and generation, with the Step-Audio-Chat version open-sourced; 2) a generative speech data engine that establishes an affordable voice cloning framework and produces the open-sourced lightweight Step-Audio-TTS-3B model through distillation; 3) an instruction-driven fine control system enabling dynamic adjustments across dialects, emotions, singing, and RAP; 4) an enhanced cognitive architecture augmented with tool calling and role-playing abilities to manage complex tasks effectively. Based on our new StepEval-Audio-360 evaluation benchmark, Step-Audio achieves state-of-the-art performance in human evaluations, especially in terms of instruction following. On open-source benchmarks like LLaMA Question, shows 9.3% average performance improvement, demonstrating our commitment to advancing the development of open-source multi-modal language technologies. Our code and models are available at https://github.com/stepfun-ai/Step-Audio.

cs.CL↗

arXiv2025

Connector-S: A Survey of Connectors in Multi-modal Large Language Models

With the rapid advancements in multi-modal large language models (MLLMs), connectors play a pivotal role in bridging diverse modalities and enhancing model performance. However, the design and evolution of connectors have not been comprehensively analyzed, leaving gaps in understanding how these components function and hindering the development of more powerful connectors. In this survey, we systematically review the current progress of connectors in MLLMs and present a structured taxonomy that categorizes connectors into atomic operations (mapping, compression, mixture of experts) and holistic designs (multi-layer, multi-encoder, multi-modal scenarios), highlighting their technical contributions and advancements. Furthermore, we discuss several promising research frontiers and challenges, including high-resolution input, dynamic compression, guide information selection, combination strategy, and interpretability. This survey is intended to serve as a foundational reference and a clear roadmap for researchers, providing valuable insights into the design and optimization of next-generation connectors to enhance the performance and adaptability of MLLMs.

cs.LG↗

arXiv2025

A Bias-Correction Decentralized Stochastic Gradient Algorithm with Momentum Acceleration

Distributed stochastic optimization algorithms can simultaneously process large-scale datasets, significantly accelerating model training. However, their effectiveness is often hindered by the sparsity of distributed networks and data heterogeneity. In this paper, we propose a momentum-accelerated distributed stochastic gradient algorithm, termed Exact-Diffusion with Momentum (EDM), which mitigates the bias from data heterogeneity and incorporates momentum techniques commonly used in deep learning to enhance convergence rate. Our theoretical analysis demonstrates that the EDM algorithm converges sub-linearly to the neighborhood of the optimal solution, the radius of which is irrespective of data heterogeneity, when applied to non-convex objective functions; under the Polyak-Lojasiewicz condition, which is a weaker assumption than strong convexity, it converges linearly to the target region. Our analysis techniques employed to handle momentum in complex distributed parameter update structures yield a sufficiently tight convergence upper bound, offering a new perspective for the theoretical analysis of other momentum-based distributed algorithms.

cs.LG↗

arXiv2025

PH-VAE: A Polynomial Hierarchical Variational Autoencoder Towards Disentangled Representation Learning

The variational autoencoder (VAE) is a simple and efficient generative artificial intelligence method for modeling complex probability distributions of various types of data, such as images and texts. However, it suffers some main shortcomings, such as lack of interpretability in the latent variables, difficulties in tuning hyperparameters while training, producing blurry, unrealistic downstream outputs or loss of information due to how it calculates loss functions and recovers data distributions, overfitting, and origin gravity effect for small data sets, among other issues. These and other limitations have caused unsatisfactory generation effects for the data with complex distributions. In this work, we proposed and developed a polynomial hierarchical variational autoencoder (PH-VAE), in which we used a polynomial hierarchical date format to generate or to reconstruct the data distributions. In doing so, we also proposed a novel Polynomial Divergence in the loss function to replace or generalize the Kullback-Leibler (KL) divergence, which results in systematic and drastic improvements in both accuracy and reproducibility of the re-constructed distribution function as well as the quality of re-constructed data images while keeping the dataset size the same but capturing fine resolution of the data. Moreover, we showed that the proposed PH-VAE has some form of disentangled representation learning ability.

cs.LG↗

arXiv2025

Bounded dissipation law and profiles of turbulent velocity moments in wall flows

Turbulent wall flows offer the most direct means for understanding the effects of boundaries and viscosity on turbulent fluctuations. Available data on mean-square fluctuations in these flows show apparent contradiction with classical scaling based on the {mean} wall shear stress. We had earlier proposed an alternative model based on the principle of bounded dissipation to describe the data. Despite its putative success, a conclusive outcome requires much higher Reynolds numbers than are available at present, or can be expected to be available in the near future. However, the model can be validated satisfactorily even within the Reynolds number range already available by considering high-order moments and their distributions in the wall-normal direction. Expressions for high-order moments of streamwise velocity fluctuation $u$ are derived in the form $ \langle u^{+2q} \rangle^{1/q}=α_q-β_q y^{\ast1/4}$; here $q$ is an integer, $α_q$ and $β_q$ are constants independent of the friction Reynolds number $Re_τ$, and $y^{\ast} = y/δ$ is the distance away from the wall, normalized by the flow thickness $δ$; in particular, $α_q =μ+σq$ according to the `linear q-norm Gaussian' process, where $μ$ and $σ$ are flow-independent constants. Excellent agreement is found between this formula and the available data in boundary layers, pipes and channels for $1 \leq q \leq 5$. For fixed $y^+ = y^*Re_τ$, the present formulation leads to the bounded state $\langle u^{+2q} \rangle^{1/q}=α_q$ as $Re_τ\rightarrow\infty$. This work demonstrates the success of the present model in describing the behavior of fluctuations in wall flows.

physics.flu-dyn↗

arXiv2025

VTG-LLM: Integrating Timestamp Knowledge into Video LLMs for Enhanced Video Temporal Grounding

Video Temporal Grounding (VTG) strives to accurately pinpoint event timestamps in a specific video using linguistic queries, significantly impacting downstream tasks like video browsing and editing. Unlike traditional task-specific models, Video Large Language Models (video LLMs) can handle multiple tasks concurrently in a zero-shot manner. Consequently, exploring the application of video LLMs for VTG tasks has become a burgeoning research area. However, despite considerable advancements in video content understanding, video LLMs often struggle to accurately pinpoint timestamps within videos, limiting their effectiveness in VTG tasks. To address this, we introduce VTG-LLM, a model designed to enhance video LLMs' timestamp localization abilities. Our approach includes: (1) effectively integrating timestamp knowledge into visual tokens; (2) incorporating absolute-time tokens to manage timestamp knowledge without concept shifts; and (3) introducing a lightweight, high-performance, slot-based token compression technique designed to accommodate the demands of a large number of frames to be sampled for VTG tasks. Additionally, we present VTG-IT-120K, a collection of publicly available VTG datasets that we have re-annotated to improve upon low-quality annotations. Our comprehensive experiments demonstrate the superior performance of VTG-LLM in comparison to other video LLM methods across a variety of VTG tasks.

cs.CV↗

arXiv2025

Heterogeneity-aware Personalized Federated Learning via Adaptive Dual-Agent Reinforcement Learning

Federated Learning (FL) empowers multiple clients to collaboratively train machine learning models without sharing local data, making it highly applicable in heterogeneous Internet of Things (IoT) environments. However, intrinsic heterogeneity in clients' model architectures and computing capabilities often results in model accuracy loss and the intractable straggler problem, which significantly impairs training effectiveness. To tackle these challenges, this paper proposes a novel Heterogeneity-aware Personalized Federated Learning method, named HAPFL, via multi-level Reinforcement Learning (RL) mechanisms. HAPFL optimizes the training process by incorporating three strategic components: 1) An RL-based heterogeneous model allocation mechanism. The parameter server employs a Proximal Policy Optimization (PPO)-based RL agent to adaptively allocate appropriately sized, differentiated models to clients based on their performance, effectively mitigating performance disparities. 2) An RL-based training intensity adjustment scheme. The parameter server leverages another PPO-based RL agent to dynamically fine-tune the training intensity for each client to further enhance training efficiency and reduce straggling latency. 3) A knowledge distillation-based mutual learning mechanism. Each client deploys both a heterogeneous local model and a homogeneous lightweight model named LiteModel, where these models undergo mutual learning through knowledge distillation. This uniform LiteModel plays a pivotal role in aggregating and sharing global knowledge, significantly enhancing the effectiveness of personalized local training. Experimental results across multiple benchmark datasets demonstrate that HAPFL not only achieves high accuracy but also substantially reduces the overall training time by 20.9%-40.4% and decreases straggling latency by 19.0%-48.0% compared to existing solutions.

cs.LG↗

arXiv2025

Completion of Lunar Magma Ocean Solidification at 4.43 Ga

Crystallization of the lunar magma ocean yielded a chemically unique liquid residuum named KREEP. This component is expressed as a large patch on the near side of the Moon, and a possible smaller patch in the northwest portion of the Moon's South Pole-Aitken basin on the far side. Thermal models estimate that the crystallization of the lunar magma ocean (LMO) could have spanned from 10 and 200 Myr, while studies of radioactive decay systems have yielded inconsistent ages for the completion of LMO crystallization covering over 160 Myr. Here, we show that the Moon achieved over 99 percent crystallization at 4429+/-76 Myr, indicating a lunar formation age of 4450 Myr or possibly older. Using the 176Lu-176Hf decay system (t1/2=37 Gyr), we found that the initial 176Hf/177Hf ratios of lunar zircons with varied U-Pb ages are consistent with their crystallization from a KREEP-rich reservoir with a consistently low 176Lu/177Hf ratio of 0.0167 that emerged ~140 Myr after solar system formation. The previously proposed younger model age of 4.33 Ga for the source of mare basalts (240 Myr after solar system formation) might reflect the timing of a large impact. Our results demonstrate that lunar magma ocean crystallization took place while the Moon was still battered by planetary embryos and planetesimals leftover from the main stage of planetary accretion. Study of Lu-Hf model ages for samples brought back from the South Pole-Aitken basin will help to assess the lateral continuity of KREEP and further understand its significance in the early history of the Moon.

astro-ph.EP↗

arXiv2025

Automorphism and Cohomology II: Complete intersections

We prove that the automorphism group of a general complete intersection $X$ in a projective space is trivial with a few well-understood exceptions. We also prove that the automorphism group of a complete intersection $X$ acts on the cohomology of $X$ faithfully with a few well-understood exceptions.

math.AG↗

arXiv2025

Stable inversion of potential in nonlinear wave equations with cubic nonlinearity

This paper investigates inverse potential problems of wave equations with cubic nonlinearity. We develop a methodology for establishing stability estimates for inversion of lower order coefficients. The new ingredients of our approach include trilinear approximations of nonlinear response operators, symbol estimates of distorted plane waves, and lower order symbol calculus.

math.AP↗

arXiv2025

Disentangled Modeling of Preferences and Social Influence for Group Recommendation

The group recommendation (GR) aims to suggest items for a group of users in social networks. Existing work typically considers individual preferences as the sole factor in aggregating group preferences. Actually, social influence is also an important factor in modeling users' contributions to the final group decision. However, existing methods either neglect the social influence of individual members or bundle preferences and social influence together as a unified representation. As a result, these models emphasize the preferences of the majority within the group rather than the actual interaction items, which we refer to as the preference bias issue in GR. Moreover, the self-supervised learning (SSL) strategies they designed to address the issue of group data sparsity fail to account for users' contextual social weights when regulating group representations, leading to suboptimal results. To tackle these issues, we propose a novel model based on Disentangled Modeling of Preferences and Social Influence for Group Recommendation (DisRec). Concretely, we first design a user-level disentangling network to disentangle the preferences and social influence of group members with separate embedding propagation schemes based on (hyper)graph convolution networks. We then introduce a socialbased contrastive learning strategy, selectively excluding user nodes based on their social importance to enhance group representations and alleviate the group-level data sparsity issue. The experimental results demonstrate that our model significantly outperforms state-of-the-art methods on two realworld datasets.

cs.IR↗

arXiv2025

MangaNinja: Line Art Colorization with Precise Reference Following

Derived from diffusion models, MangaNinjia specializes in the task of reference-guided line art colorization. We incorporate two thoughtful designs to ensure precise character detail transcription, including a patch shuffling module to facilitate correspondence learning between the reference color image and the target line art, and a point-driven control scheme to enable fine-grained color matching. Experiments on a self-collected benchmark demonstrate the superiority of our model over current solutions in terms of precise colorization. We further showcase the potential of the proposed interactive point control in handling challenging cases, cross-character colorization, multi-reference harmonization, beyond the reach of existing algorithms.

cs.CV↗

arXiv2025

Guided modes in graphene waveguides

By analogy of optical waveguides, we investigate the guided modes in graphene waveguides, which is made of symmetric quantum well. The unique properties of the graphene waveguide are discussed based on the two different dispersion relations, which correspond to classical motion and Klein tunneling, respectively. It is shown that the third-order mode is absent in the classical motion, while the fundamental mode is absent in the Klein tunneling case. We hope these phenomena can lead to the potential applications in graphene-based quantum devices.

cond-mat.mes-hall↗

arXiv2025

Tunable lateral displacement and spin beam splitter for ballistic electrons in two-dimensional magnetic-electric nanostructures

We investigate the lateral displacements for ballistic electron beams in a two-dimensional electron gas modulated by metallic ferromagnetic (FM) stripes with parallel and antiparallel (AP) magnetization configurations. It is shown that the displacements are negative as well as positive, which can be controlled by adjusting the electric potential induced by the applied voltage and the magnetic field strength of FM stripes. Based on these phenomena, we propose an efficient way to realize a spin beam splitter, which can completely separate spin-up and spin-down electron beams in the AP configuration by their corresponding spatial positions.

cond-mat.mes-hall↗

arXiv2025

Comprehensive Examination of Unrolled Networks for Solving Linear Inverse Problems

Unrolled networks have become prevalent in various computer vision and imaging tasks. Although they have demonstrated remarkable efficacy in solving specific computer vision and computational imaging tasks, their adaptation to other applications presents considerable challenges. This is primarily due to the multitude of design decisions that practitioners working on new applications must navigate, each potentially affecting the network's overall performance. These decisions include selecting the optimization algorithm, defining the loss function, and determining the number of convolutional layers, among others. Compounding the issue, evaluating each design choice requires time-consuming simulations to train, fine-tune the neural network, and optimize for its performance. As a result, the process of exploring multiple options and identifying the optimal configuration becomes time-consuming and computationally demanding. The main objectives of this paper are (1) to unify some ideas and methodologies used in unrolled networks to reduce the number of design choices a user has to make, and (2) to report a comprehensive ablation study to discuss the impact of each of the choices involved in designing unrolled networks and present practical recommendations based on our findings. We anticipate that this study will help scientists and engineers design unrolled networks for their applications and diagnose problems within their networks efficiently.

eess.IV↗

arXiv2025

Terabit-class coherent communications enabled by an integrated photonics erbium doped amplifier

Coherent technologies have revolutionized optical communications, driving the capacity per fiber to multi-terabit per second (Tb/s) in combination with wavelength division multiplexing (WDM). With an ever-increasing deployment density of coherent systems, the demand for highly integrated WDM coherent transceivers has been rising. While tremendous progress has been made on silicon photonics compatible high-speed modulation and photodetection on chip, a solution for monolithically integrable amplifier with high gain and output power remains a challenge. Recently, an erbium doped waveguide amplifier based on ultra-low loss silicon nitride waveguides has demonstrated gain and output power levels potentially suitable for Terabit class coherent communications. Here, we demonstrate a WDM coherent system enabled by this integrated photonic amplification solution. The system uses the waveguide amplifier as a booster amplifier of 16 WDM signals each carrying a net data rate of 1.6 Tb/s, achieving 25.6-Tb/s net capacity over 81-km fiber transmission. Our results highlight a fully integrated solution for highly parallel coherent transceivers including amplification, that has the potential to transform future optical communications.

physics.optics↗

arXiv2025

Embodied VideoAgent: Persistent Memory from Egocentric Videos and Embodied Sensors Enables Dynamic Scene Understanding

This paper investigates the problem of understanding dynamic 3D scenes from egocentric observations, a key challenge in robotics and embodied AI. Unlike prior studies that explored this as long-form video understanding and utilized egocentric video only, we instead propose an LLM-based agent, Embodied VideoAgent, which constructs scene memory from both egocentric video and embodied sensory inputs (e.g. depth and pose sensing). We further introduce a VLM-based approach to automatically update the memory when actions or activities over objects are perceived. Embodied VideoAgent attains significant advantages over counterparts in challenging reasoning and planning tasks in 3D scenes, achieving gains of 4.9% on Ego4D-VQ3D, 5.8% on OpenEQA, and 11.7% on EnvQA. We have also demonstrated its potential in various embodied AI tasks including generating embodied interactions and perception for robot manipulation. The code and demo will be made public.

cs.CV↗