arXiv subjects

Hongli Wang

Publications and source records attributed to Hongli Wang.

12 recordsLinked to original sources

arXiv2026

SenseNova-U1.5: Towards Native Unified Visual Intelligence

We launch SenseNova-U1.5, an 8B-MoT native unified multimodal model that understands, reasons about, and generates visual content within an encoder-free and VAE-free architecture. We strengthen its visual interface through spatially coherent patch reconstruction and scale its training with carefully curated generation and editing data, improved task formulation, structural prompt enhancement, and native resolutions of up to 4K. For post-training, we optimize specialized experts for visual aesthetics, bilingual text rendering, infographic generation, and image editing, and consolidate their capabilities through multi-expert on-policy distillation. Across extensive evaluations, SenseNova-U1.5 largely advances image fidelity, text rendering, complex composition, multi-reference editing, and interleaved generation, while improving instruction following and preserving subject identity, geometry, and unmodified regions. Despite limited exposure to structured formats in its generation data, SenseNova-U1.5 generalizes effectively to long, complex, and structured visual instructions, further proving that multimodal understanding can transfer to visual planning and creation. Together, these findings position native unified modelling as a promising path towards systems that perceive, reason and create within a fully end-to-end framework. We will open-source training code, including supervised fine-tuning, reinforcement learning, and on-policy distillation.

cs.CV↗

arXiv2026

SenseNova-U1: Unifying Multimodal Understanding and Generation with NEO-unify Architecture

Recent large vision-language models (VLMs) remain fundamentally constrained by a persistent dichotomy: understanding and generation are treated as distinct problems, leading to fragmented architectures, cascaded pipelines, and misaligned representation spaces. We argue that this divide is not merely an engineering artifact, but a structural limitation that hinders the emergence of native multimodal intelligence. Hence, we introduce SenseNova-U1, a native unified multimodal paradigm built upon NEO-unify, in which understanding and generation evolve as synergistic views of a single underlying process. We launch two native unified variants, SenseNova-U1-8B-MoT and SenseNova-U1-A3B-MoT, built on dense (8B) and mixture-of-experts (30B-A3B) understanding baselines, respectively. Designed from first principles, they rival top-tier understanding-only VLMs across text understanding, vision-language perception, knowledge reasoning, agentic decision-making, and spatial intelligence. Meanwhile, they deliver strong semantic consistency and visual fidelity, excelling in conventional or knowledge-intensive any-to-image (X2I) synthesis, complex text-rich infographic generation, and interleaved vision-language generation, with or without think patterns. Beyond performance, we show detailed model design, data preprocessing, pre-/post-training, and inference strategies to support community research. Last but not least, preliminary evidence demonstrates that our models extend beyond perception and generation, performing strongly in vision-language-action (VLA) and world model (WM) scenarios. This points toward a broader roadmap where models do not translate between modalities, but think and act across them in a native manner. Multimodal AI is no longer about connecting separate systems, but about building a unified one and trusting the necessary capabilities to emerge from within.

cs.CV↗

arXiv2025

Upper-semicontinuity of uniform attractors for the non-autonomous viscoelastic Kirchhoff plate equation with memory

This paper delves into the long-time dynamics of a non-autonomous viscoelastic Kirchhoff plate equation with memory effects, described by $$ u_{t t}-\Delta u_{t t}+a_\epsilon(t) u_t+\alpha \Delta^2 u-\int_0^{\infty} \mu(s) \Delta^2 u(t-s) \mathrm{d} s-\Delta u_t+f(u)=g(x,t), $$ in bounded domain $\Omega \subset \mathbb{R}^N$ with smooth boundary and nonlinear terms. Initially, the global existence of a weak solution that induces a continuous process is established. Subsequently, the existence of a uniform attractor is demonstrated in both subcritical and critical growth scenarios, utilizing operator techniques and an innovative analytical approach. Finally, the upper semicontinuity of the family of uniform attractors as the pert parameterurbation $\epsilon \to 0^+$ is proven through delicate energy estimates and a contradiction argument. Our results not only extend classical attractor theory to more general non-autonomous viscoelastic systems but also resolve open questions regarding the limiting behavior of attractors in the presence of both memory and critical nonlinearity.

math.AP↗

arXiv2024

Uniform attractor of a non-autonomous Lame thermoelastic system

In this paper, we investigate the dynamical behavior of non-autonomous Lame thermoelastic systems within $N$-dimensional materials. With appropriate constraints on nonlinear characteristics and functional parameters, we initially establish the existence of a uniformly absorbing set by constructing a Lyapunov function. Subsequently, we employ the contraction mapping principle to demonstrate the uniformly asymptotic compactness of the system. Finally, under irrotational conditions, we prove the existence of a uniform attractor $\mathcal{A}_\Sigma$ in the space $H_c$.

math.AP↗

arXiv2024

Static Deep Q-learning for Green Downlink C-RAN

Power saving is a main pillar in the operation of wireless communication systems. In this paper, we investigate cloud radio access network (C-RAN) capability to reduce power consumption based on the user equipment (UE) requirement. Aiming to save the long-term C-RAN energy consumption, an optimization problem is formulated to manage the downlink power without degrading the UE requirement by designing the power offset parameter. Considering stochastic traffic arrivals at UEs, we first formulate the problem as a Markov decision process (MDP) and then set up a dual objective optimization problem in terms of the downlink throughput and power. To solve this optimization problem, we develop a novel static deep Q-learning (SDQL) algorithm to maximize the downlink throughput and minimize the downlink power. In our proposed algorithm, we design multi-Q-tables to simultaneously optimize power reductions of activated RRHs by assigning one Q-table for each UE. To maximize the accumulative reward in terms of the downlink throughput loss and power reduction, our proposed algorithm performs power reductions of activated RRHs through continuous environmental interactions. Simulation results1 show that our proposed algorithm enjoys a superior average power reduction compared to the activation and sleep schemes, and enjoys a low computational complexity.

cs.IT↗

arXiv2023

AdaMEC: Towards a Context-Adaptive and Dynamically-Combinable DNN Deployment Framework for Mobile Edge Computing

With the rapid development of deep learning, recent research on intelligent and interactive mobile applications (e.g., health monitoring, speech recognition) has attracted extensive attention. And these applications necessitate the mobile edge computing scheme, i.e., offloading partial computation from mobile devices to edge devices for inference acceleration and transmission load reduction. The current practices have relied on collaborative DNN partition and offloading to satisfy the predefined latency requirements, which is intractable to adapt to the dynamic deployment context at runtime. AdaMEC, a context-adaptive and dynamically-combinable DNN deployment framework is proposed to meet these requirements for mobile edge computing, which consists of three novel techniques. First, once-for-all DNN pre-partition divides DNN at the primitive operator level and stores partitioned modules into executable files, defined as pre-partitioned DNN atoms. Second, context-adaptive DNN atom combination and offloading introduces a graph-based decision algorithm to quickly search the suitable combination of atoms and adaptively make the offloading plan under dynamic deployment contexts. Third, runtime latency predictor provides timely latency feedback for DNN deployment considering both DNN configurations and dynamic contexts. Extensive experiments demonstrate that AdaMEC outperforms state-of-the-art baselines in terms of latency reduction by up to 62.14% and average memory saving by 55.21%.

cs.DC↗

arXiv2021

Nanoscale Non-Destructive Ferroelectric Characterization with Non-Contact Heterodyne Electrostrain Force Microscopy

Perceiving nanoscale ferroelectric phenomena from real space is of great importance for elucidating underlying ferroelectric physics. During the past decades, nanoscale ferroelectric characterization has mainly relied on the Piezoresponse Force Microscopy (PFM), however, the fundamental limitations of PFM have made the nanoscale ferroelectric studies encounter significant bottlenecks. In this study, a high-resolution non-contact ferroelectric measurement, named Non-Contact Heterodyne Electrostrain Force Microscopy (NC-HEsFM), has been introduced firstly. It has been unambiguously demonstrated that NC-HEsFM can operate on multiple eigenmodes to perform ideal high-resolution ferroelectric domain mapping, standard ferroelectric hysteresis loop measurement and controllable domain manipulation. With using quartz tuning fork (QTF) sensor and heterodyne detection, NC-HEsFM shows an unprecedented capability in achieving real non-contact yet non-destructive ferroelectric characterization with negligible electrostatic force effect. It is believed that NC-HEsFM can be extensively used in various ferroelectric or piezoelectric studies with providing substantially improved characterization performance. Meanwhile, the QTF-based force detection makes NC-HEsFM highly compatible for high-vacuum and low-temperature environments, providing ideal conditions for achieving an ultra-high spatial resolution to investigate the most intrinsic ferroelectric phenomena.

cond-mat.mtrl-sci↗

arXiv2020

A New Heterodyne Megasonic Piezoresponse Force Microscopy with High-frequency Excitation beyond 100 MHz

Piezoresponse Force Microscopy (PFM), as a powerful nanoscale characterization technique, has been extensively utilized to elucidate diverse underlying physics of ferroelectricity. However, the intensive study of conventional PFM has revealed a growing number of concerns and limitations which are largely challenging its validity and application. Herein, we developed a new advanced PFM technique, named Heterodyne Megasonic Piezoresponse Force Microscopy (HM-PFM), which uniquely uses 106 to 108 Hz high-frequency excitation and heterodyne method to measure the piezoelectric strain at nanoscale. We report that HM-PFM can unambiguously provide standard ferroelectric domain and hysteresis loop measurements, and an effective domain characterization with excitation frequency up to ~110 MHz has been realized. Most importantly, owing to the high-frequency and heterodyne scheme, the contributions from both electrostatic force and electrochemical strain can be significantly minimized in HM-PFM. Furthermore, a special difference-frequency piezoresponse frequency spectrum (DFPFS) measurement is developed on HM-PFM and a distinct DFPFS characteristic is observed on the materials with piezoelectricity. It is believed that HM-PFM can be an excellent candidate for the piezoelectric or ferroelectric studies where the conventional PFM results are highly controversial.

physics.app-ph↗

arXiv2018

Variation of Contact Resonance Frequency during Domain Switching in PFM Measurements for Ferroelectric Materials

Piezoresponse Force Spectroscopy (PFS) is a powerful method widely used for measuring the nanoscale ferroelectric responses of the materials. However, it is found that certain non-ferroelectric materials can also generate similar responses from the PFS measurements due to many other factors, hence, it is believed that PFS alone is not sufficient to differentiate the ferroelectric and non-ferroelectric materials. On the other hands, this work shows that there are distinct differences in contact resonance frequency variation during the PFS measurements for ferroelectric and non-ferroelectric materials. Therefore, a new, simple and effective method is proposed to differentiate the responses from the ferroelectric and non-ferroelectric materials, this new analysis uses contact resonance frequency responses during the PFS measurements as a new parameter to differentiate the PFS measured responses from different materials.

cond-mat.mtrl-sci↗

arXiv2018

Inviscid, incompressible and semiclassical limits of Quantum Navier-Stokes equation

IIn the paper, we consider the inviscid, incompressible and semiclassical limits limits of the barotropic quantum Navier-Stokes equations of compressible flows in a periodic domain. We show that the limit solutions satisfy the incompressible Euler system based on the relative entropy inequality and on the detailed analysis for general initial data. The rate of convergence is estimated in terms of the Mach number.

math.AP↗

arXiv2015

The free energy cost of accurate biochemical oscillations

Oscillation is an important cellular process that regulates timing of different vital life cycles. However, in the noisy cellular environment, oscillations can be highly inaccurate due to phase fluctuations. It remains poorly understood how biochemical circuits suppress phase fluctuations and what is the incurred thermodynamic cost. Here, we study four different types of biochemical oscillations representing three basic oscillation motifs shared by all known oscillatory systems. We find that the phase diffusion constant follows the same inverse dependence on the free energy dissipation per period for all systems studied. This relationship between the phase diffusion and energy dissipation is shown analytically in a model of noisy oscillation. Microscopically, we find that the oscillation is driven by multiple irreversible cycles that hydrolyze the fuel molecules such as ATP; the number of phase coherent periods is proportional to the free energy consumed per period. Experimental evidence in support of this universal relationship and testable predictions are also presented.

physics.bio-ph↗

arXiv2012

Dynamical topology and statistical properties of spatiotemporal chaos

For spatiotemporal chaos described by partial differential equations, there are generally locations where the dynamical variable achieves its local extremum or where the time partial derivative of the variable vanishes instantaneously. To a large extent, the location and movement of these topologically special points determine the qualitative structure of the disordered states. We analyze numerically statistical properties of the topologically special points in one-dimensional spatiotemporal chaos. The probability distribution functions for the number of point, the lifespan, and the distance covered during their lifetime are obtained from numerical simulations. Mathematically, we establish a probabilistic model to describe the dynamics of these topologically special points. In despite of the different definitions in different spatiotemporal chaos, the dynamics of these special points can be described in a uniform approach.

nlin.CD↗