arXiv subjects

Chen Zhang

Publications and source records attributed to Chen Zhang.

At least 217 records · Page 12Linked to original sources

arXiv2025

The soft X-ray transient EP241021a: A cosmic explosion with a complex off-axis jet and cocoon from a massive progenitor

X-ray flashes (XRFs) are fast X-ray transients thought to be softer analogs of gamma-ray bursts (GRBs). With its soft X-ray sensitivity, the Einstein Probe (EP) provides a unique opportunity to study these events. We report multiwavelength observations of EP241021a, a soft X-ray transient detected by EP, and interpret its afterglow in the context of leading XRF models. The prompt emission was observed by EP-WXT and Fermi-GBM, followed by a broad campaign across radio (uGMRT, ATCA, e-MERLIN, ALMA), optical (LBT, GTC, CAHA), and X-rays (EP-FXT). Light curves and spectra were analyzed with both empirical and physical models of GRBs and spherical expansions (both nonrelativistic and mildly relativistic cocoons). The afterglow shows multiple components, consistent with a structured jet interacting with a complex environment. The early optical and X-ray decline is explained by wide, low-Lorentz-factor ($γ\sim 40$) wings, while a rebrightening at approximately 7 days arises from the off-axis jet core. Radio data require an additional mildly relativistic cocoon ($γ\sim 2$), and a late (70 days) spectral component peaking at 50 GHz suggests a second, slower cocoon ($γ\sim 1$).

astro-ph.HE↗

arXiv2025

Manipulating terahertz phonon-polariton in the ultrastrong coupling regime with bound states in the continuum

The strong coupling between photons and phonons in polar materials gives rise to phonon-polaritons that encapsulate a wealth of physical information, offering crucial tools for the ultrafast terahertz sources and the topological engineering of terahertz light. However, it is still quite challenging to form and manipulate the terahertz phonon-polaritons under the ultrastrong coupling regime till now. In this work, we demonstrate the ultrastrong coupling between the phonon (at 0.95 THz) in a MaPbI 3 film and the metallic bound states in the continuum (BICs) in Au metasurfaces. The Rabi splitting can be continuously tuned from 28% to 48.4% of the phonon frequency by adjusting the parameters (size, shape and period) of Au metasurfaces, reaching the ultrastrong coupling regime. By introducing wavelet transform, the mode evolution information of the terahertz phonon-polariton is successfully extracted. It indicates that the phonon radiation intensity of the MaPbI 3 film is enhanced as the coupling strength is increased. This work not only establishes a new platform for terahertz devices but also opens new avenues for exploring the intricate dynamics of terahertz phonon-polaritons.

physics.optics↗

arXiv2025

LongLLaVA: Scaling Multi-modal LLMs to 1000 Images Efficiently via a Hybrid Architecture

Expanding the long-context capabilities of Multi-modal Large Language Models~(MLLMs) is critical for advancing video understanding and high-resolution image analysis. Achieving this requires systematic improvements in model architecture, data construction, and training strategies, particularly to address challenges such as performance degradation with increasing image counts and high computational costs. In this paper, we propose a hybrid architecture that integrates Mamba and Transformer blocks, introduce data construction methods that capture both temporal and spatial dependencies, and employ a progressive training strategy. Our released model, LongLLaVA (\textbf{Long}-Context \textbf{L}arge \textbf{L}anguage \textbf{a}nd \textbf{V}ision \textbf{A}ssistant), demonstrates an effective balance between efficiency and performance. LongLLaVA achieves competitive results across various benchmarks while maintaining high throughput and low memory consumption. Notably, it can process nearly one thousand images on a single A100 80GB GPU, underscoring its potential for a wide range of multi-modal applications.

cs.CL↗

arXiv2025

LongCat-Flash Technical Report

We introduce LongCat-Flash, a 560-billion-parameter Mixture-of-Experts (MoE) language model designed for both computational efficiency and advanced agentic capabilities. Stemming from the need for scalable efficiency, LongCat-Flash adopts two novel designs: (a) Zero-computation Experts, which enables dynamic computational budget allocation and activates 18.6B-31.3B (27B on average) per token depending on contextual demands, optimizing resource usage. (b) Shortcut-connected MoE, which enlarges the computation-communication overlap window, demonstrating notable gains in inference efficiency and throughput compared to models of a comparable scale. We develop a comprehensive scaling framework for large models that combines hyperparameter transfer, model-growth initialization, a multi-pronged stability suite, and deterministic computation to achieve stable and reproducible training. Notably, leveraging the synergy among scalable architectural design and infrastructure efforts, we complete model training on more than 20 trillion tokens within 30 days, while achieving over 100 tokens per second (TPS) for inference at a cost of \$0.70 per million output tokens. To cultivate LongCat-Flash towards agentic intelligence, we conduct a large-scale pre-training on optimized mixtures, followed by targeted mid- and post-training on reasoning, code, and instructions, with further augmentation from synthetic data and tool use tasks. Comprehensive evaluations demonstrate that, as a non-thinking foundation model, LongCat-Flash delivers highly competitive performance among other leading models, with exceptional strengths in agentic tasks. The model checkpoint of LongCat-Flash is open-sourced to foster community research. LongCat Chat: https://longcat.ai Hugging Face: https://huggingface.co/meituan-longcat GitHub: https://github.com/meituan-longcat

cs.CL↗

arXiv2025

AC-Refiner: Efficient Arithmetic Circuit Optimization Using Conditional Diffusion Models

Arithmetic circuits, such as adders and multipliers, are fundamental components of digital systems, directly impacting the performance, power efficiency, and area footprint. However, optimizing these circuits remains challenging due to the vast design space and complex physical constraints. While recent deep learning-based approaches have shown promise, they struggle to consistently explore high-potential design variants, limiting their optimization efficiency. To address this challenge, we propose AC-Refiner, a novel arithmetic circuit optimization framework leveraging conditional diffusion models. Our key insight is to reframe arithmetic circuit synthesis as a conditional image generation task. By carefully conditioning the denoising diffusion process on target quality-of-results (QoRs), AC-Refiner consistently produces high-quality circuit designs. Furthermore, the explored designs are used to fine-tune the diffusion model, which focuses the exploration near the Pareto frontier. Experimental results demonstrate that AC-Refiner generates designs with superior Pareto optimality, outperforming state-of-the-art baselines. The performance gain is further validated by integrating AC-Refiner into practical applications.

cs.AR↗

arXiv2025

Bipedal Balance Control with Whole-body Musculoskeletal Standing and Falling Simulations

Balance control is important for human and bipedal robotic systems. While dynamic balance during locomotion has received considerable attention, quantitative understanding of static balance and falling remains limited. This work presents a hierarchical control pipeline for simulating human balance via a comprehensive whole-body musculoskeletal system. We identified spatiotemporal dynamics of balancing during stable standing, revealed the impact of muscle injury on balancing behavior, and generated fall contact patterns that aligned with clinical data. Furthermore, our simulated hip exoskeleton assistance demonstrated improvement in balance maintenance and reduced muscle effort under perturbation. This work offers unique muscle-level insights into human balance dynamics that are challenging to capture experimentally. It could provide a foundation for developing targeted interventions for individuals with balance impairments and support the advancement of humanoid robotic systems.

cs.RO↗

arXiv2025

Optimal Estimation for General Gaussian Processes

This paper proposes a novel exact maximum likelihood (ML) estimation method for general Gaussian processes, where all parameters are estimated jointly. The exact ML estimator (MLE) is consistent and asymptotically normally distributed. We prove the local asymptotic normality (LAN) property of the sequence of statistical experiments for general Gaussian processes in the sense of Le Cam, thereby enabling optimal estimation and facilitating statistical inference. The results rely solely on the asymptotic behavior of the spectral density near zero, allowing them to be widely applied. The established optimality not only addresses the gap left by Adenstedt(1974), who proposed an efficient but infeasible estimator for the long-run mean $μ$, but also enables us to evaluate the finite-sample performance of the existing method -- the commonly used plug-in MLE, in which the sample mean is substituted into the likelihood. Our simulation results show that the plug-in MLE performs nearly as well as the exact MLE, alleviating concerns that inefficient estimation of $μ$ would compromise the efficiency of the remaining parameter estimates.

math.ST↗

arXiv2025

Brillouin Platycosms and Topological Phases

There exist ten distinct closed flat $3$D manifolds, known as platycosms, which hold significance in mathematics and have been postulated as potential geometric models for our universe. In this work, we demonstrate their manifestation as universes of Bloch particles, namely as momentum-space units referred to as Brillouin platycosms, which are natural extensions of the Brillouin torus within a broader framework of projective crystallographic symmetries. Moreover, we provide exact K-theoretical classifications of topological insulators over these platycosms by the Atiyah-Hirzebruch spectral sequence, and formulate a complete set of topological invariants for their identification. Topological phase transitions are generically characterized by Weyl semimetals, adhering to the generalized Nielsen-Ninomiya theorem: the total chirality number over a Brillouin platycosm is even (zero) if the platycosm is non-orientable (orientable). Our work generalizes the notion of Brillouin torus to ten Brillouin platycosms and therefore fundamentally diversifies the stages on which Block wavefunctions can perform their topological dance.

cond-mat.mes-hall↗

arXiv2025

Apple Intelligence Foundation Language Models: Tech Report 2025

We introduce two multilingual, multimodal foundation language models that power Apple Intelligence features across Apple devices and services: i a 3B-parameter on-device model optimized for Apple silicon through architectural innovations such as KV-cache sharing and 2-bit quantization-aware training; and ii a scalable server model built on a novel Parallel-Track Mixture-of-Experts PT-MoE transformer that combines track parallelism, mixture-of-experts sparse computation, and interleaved global-local attention to deliver high quality with competitive cost on Apple's Private Cloud Compute platform. Both models are trained on large-scale multilingual and multimodal datasets sourced via responsible web crawling, licensed corpora, and high-quality synthetic data, then further refined with supervised fine-tuning and reinforcement learning on a new asynchronous platform. The resulting models support several additional languages while understanding images and executing tool calls. In public benchmarks and human evaluations, both the server model and the on-device model match or surpass comparably sized open baselines. A new Swift-centric Foundation Models framework exposes guided generation, constrained tool calling, and LoRA adapter fine-tuning, allowing developers to integrate these capabilities with a few lines of code. The latest advancements in Apple Intelligence models are grounded in our Responsible AI approach with safeguards like content filtering and locale-specific evaluation, as well as our commitment to protecting our users' privacy with innovations like Private Cloud Compute.

cs.LG↗

arXiv2025

InfinityHuman: Towards Long-Term Audio-Driven Human

Audio-driven human animation has attracted wide attention thanks to its practical applications. However, critical challenges remain in generating high-resolution, long-duration videos with consistent appearance and natural hand motions. Existing methods extend videos using overlapping motion frames but suffer from error accumulation, leading to identity drift, color shifts, and scene instability. Additionally, hand movements are poorly modeled, resulting in noticeable distortions and misalignment with the audio. In this work, we propose InfinityHuman, a coarse-to-fine framework that first generates audio-synchronized representations, then progressively refines them into high-resolution, long-duration videos using a pose-guided refiner. Since pose sequences are decoupled from appearance and resist temporal degradation, our pose-guided refiner employs stable poses and the initial frame as a visual anchor to reduce drift and improve lip synchronization. Moreover, to enhance semantic accuracy and gesture realism, we introduce a hand-specific reward mechanism trained with high-quality hand motion data. Experiments on the EMTD and HDTF datasets show that InfinityHuman achieves state-of-the-art performance in video quality, identity preservation, hand accuracy, and lip-sync. Ablation studies further confirm the effectiveness of each module. Code will be made public.

cs.CV↗

arXiv2025

ClusterFusion: Expanding Operator Fusion Scope for LLM Inference via Cluster-Level Collective Primitive

Large language model (LLM) decoding suffers from high latency due to fragmented execution across operators and heavy reliance on off-chip memory for data exchange and reduction. This execution model limits opportunities for fusion and incurs significant memory traffic and kernel launch overhead. While modern architectures such as NVIDIA Hopper provide distributed shared memory and low-latency intra-cluster interconnects, they expose only low-level data movement instructions, lacking structured abstractions for collective on-chip communication. To bridge this software-hardware gap, we introduce two cluster-level communication primitives, ClusterReduce and ClusterGather, which abstract common communication patterns and enable structured, high-speed data exchange and reduction between thread blocks within a cluster, allowing intermediate results to be on-chip without involving off-chip memory. Building on these abstractions, we design ClusterFusion, an execution framework that schedules communication and computation jointly to expand operator fusion scope by composing decoding stages such as QKV Projection, Attention, and Output Projection into a single fused kernels. Evaluations on H100 GPUs show that ClusterFusion outperforms state-of-the-art inference frameworks by 1.61x on average in end-to-end latency across different models and configurations. The source code is available at https://github.com/xinhao-luo/ClusterFusion.

cs.DC↗

arXiv2025

Few-shot Unknown Class Discovery of Hyperspectral Images with Prototype Learning and Clustering

Open-set few-shot hyperspectral image (HSI) classification aims to classify image pixels by using few labeled pixels per class, where the pixels to be classified may be not all from the classes that have been seen. To address the open-set HSI classification challenge, current methods focus mainly on distinguishing the unknown class samples from the known class samples and rejecting them to increase the accuracy of identifying known class samples. They fails to further identify or discovery the unknow classes among the samples. This paper proposes a prototype learning and clustering method for discoverying unknown classes in HSIs under the few-shot environment. Using few labeled samples, it strives to develop the ability of infering the prototypes of unknown classes while distinguishing unknown classes from known classes. Once the unknown class samples are rejected by the learned known class classifier, the proposed method can further cluster the unknown class samples into different classes according to their distance to the inferred unknown class prototypes. Compared to existing state-of-the-art methods, extensive experiments on four benchmark HSI datasets demonstrate that our proposed method exhibits competitive performance in open-set few-shot HSI classification tasks. All the codes are available at \href{https://github.com/KOBEN-ff/OpenFUCD-main} {https://github.com/KOBEN-ff/OpenFUCD-main}

cs.CV↗

arXiv2025

XRISM reveals a variable, multi-phase outflow-inflow structure during the X-ray obscured 2024 outburst of the black hole transient V4641 Sgr

We report the results of a simultaneous X-ray and optical spectroscopy campaign on the Galactic black hole X-ray binary V4641 Sgr, carried out with XRISM and the Seimei telescope during a low-luminosity phase towards the end of its 2024 outburst. Despite a very low X-ray luminosity of $10^{34}$ erg s$^{-1}$, the continuum spectrum is well reproduced by a disk blackbody model with a high inner disk temperature ($1.8$ keV). XRISM/Resolve provides the highest-resolution X-ray spectrum ever obtained from the source, and several strong, narrow emission lines were detected, resolved and characterized at a high significance level. The continuum shape and narrow emission lines both indicate that the inner disk region is obscured by the surrounding high-density gas, and the intrinsic luminosity is several orders of magnitude higher. In the simultaneous optical observation from the Seimei telescope, the line features are largely dominated by the optical companion. Although we detect a clear emission component in H$α$ that could originate from a cold outflow or the disk atmosphere, there are no signs of the strong outflow signatures historically detected in this source. In X-rays, the combination of significantly redshifted ($\sim 700$ km s$^{-1}$) and weakly blueshifted ($\sim-250$ km s$^{-1}$) components, all varying strongly on ks timescales, along with a marginally significant (99.2%) highly blueshifted ($\sim-1200$ km s$^{-1}$) component, indicates a complex, inhomogeneous outflow geometry. This is corroborated by the erratic long-term evolution of the source seen in the complementary X-ray monitoring, and radio detections spanning 3 orders of magnitude.

astro-ph.HE↗

arXiv2025

MINR: Efficient Implicit Neural Representations for Multi-Image Encoding

Implicit Neural Representations (INRs) aim to parameterize discrete signals through implicit continuous functions. However, formulating each image with a separate neural network~(typically, a Multi-Layer Perceptron (MLP)) leads to computational and storage inefficiencies when encoding multi-images. To address this issue, we propose MINR, sharing specific layers to encode multi-image efficiently. We first compare the layer-wise weight distributions for several trained INRs and find that corresponding intermediate layers follow highly similar distribution patterns. Motivated by this, we share these intermediate layers across multiple images while preserving the input and output layers as input-specific. In addition, we design an extra novel projection layer for each image to capture its unique features. Experimental results on image reconstruction and super-resolution tasks demonstrate that MINR can save up to 60\% parameters while maintaining comparable performance. Particularly, MINR scales effectively to handle 100 images, maintaining an average peak signal-to-noise ratio (PSNR) of 34 dB. Further analysis of various backbones proves the robustness of the proposed MINR.

cs.CV↗

arXiv2025

Enhancing Robustness of Implicit Neural Representations Against Weight Perturbations

Implicit Neural Representations (INRs) encode discrete signals in a continuous manner using neural networks, demonstrating significant value across various multimedia applications. However, the vulnerability of INRs presents a critical challenge for their real-world deployments, as the network weights might be subjected to unavoidable perturbations. In this work, we investigate the robustness of INRs for the first time and find that even minor perturbations can lead to substantial performance degradation in the quality of signal reconstruction. To mitigate this issue, we formulate the robustness problem in INRs by minimizing the difference between loss with and without weight perturbations. Furthermore, we derive a novel robust loss function to regulate the gradient of the reconstruction loss with respect to weights, thereby enhancing the robustness. Extensive experiments on reconstruction tasks across multiple modalities demonstrate that our method achieves up to a 7.5~dB improvement in peak signal-to-noise ratio (PSNR) values compared to original INRs under noisy conditions.

cs.CV↗

arXiv2025

ZigzagAttention: Efficient Long-Context Inference with Exclusive Retrieval and Streaming Heads

With the rapid development of large language models (LLMs), handling long context has become one of the vital abilities in LLMs. Such long-context ability is accompanied by difficulties in deployment, especially due to the increased consumption of KV cache. There is certain work aiming to optimize the memory footprint of KV cache, inspired by the observation that attention heads can be categorized into retrieval heads that are of great significance and streaming heads that are of less significance. Typically, identifying the streaming heads and and waiving the KV cache in the streaming heads would largely reduce the overhead without hurting the performance that much. However, since employing both retrieval and streaming heads in one layer decomposes one large round of attention computation into two small ones, it may unexpectedly bring extra latency on accessing and indexing tensors. Based on this intuition, we impose an important improvement to the identification process of retrieval and streaming heads, in which we design a criterion that enforces exclusively retrieval or streaming heads gathered in one unique layer. In this way, we further eliminate the extra latency and only incur negligible performance degradation. Our method named \textsc{ZigzagAttention} is competitive among considered baselines owing to reduced latency and comparable performance.

cs.CL↗

arXiv2025

Contact-Rich and Deformable Foot Modeling for Locomotion Control of the Human Musculoskeletal System

The human foot serves as the critical interface between the body and environment during locomotion. Existing musculoskeletal models typically oversimplify foot-ground contact mechanics, limiting their ability to accurately simulate human gait dynamics. We developed a novel contact-rich and deformable model of the human foot integrated within a complete musculoskeletal system that captures the complex biomechanical interactions during walking. To overcome the control challenges inherent in modeling multi-point contacts and deformable material, we developed a two-stage policy training strategy to learn natural walking patterns for this interface-enhanced model. Comparative analysis between our approach and conventional rigid musculoskeletal models demonstrated improvements in kinematic, kinetic, and gait stability metrics. Validation against human subject data confirmed that our simulation closely reproduced real-world biomechanical measurements. This work advances contact-rich interface modeling for human musculoskeletal systems and establishes a robust framework that can be extended to humanoid robotics applications requiring precise foot-ground interaction control.

cs.RO↗

arXiv2025

Rethinking the Sioux Falls Network: Insights from Path-Driven Higher-Order Network Analysis

Benchmark scenarios are widely used in transportation research to evaluate routing algorithms, simulate infrastructure interventions, and test new technologies under controlled conditions. However, the structural and behavioral fidelity of these benchmarks remains largely unquantified, raising concerns about the external validity of simulation results. In this study, we introduce a mathematical framework based on higher-order network models to evaluate the representativeness of benchmark networks, focusing on the widely used Sioux Falls scenario. Higher-order network models encode empirical and simulated trajectory data into memory-aware network representations, which we use to quantify sequential dependencies in mobility behavior and assess how well benchmark networks capture real-world structural and functional patterns. Applying this framework to the Sioux Falls network, as well as real-world trajectory data, we quantify structural complexity, optimal memory length, link prediction accuracy, and centrality alignment. Our results show and statistically quantify that the classical Sioux Falls network exhibits limited path diversity, rapid structural fragmentation at higher orders, and weak alignment with empirical routing behavior. These results illustrate the potential of higher-order network models to bridge the gap between simulation-based and real-world mobility analysis, providing a robust foundation for more accurate and generalizable insights in transportation research.

cs.CE↗