arXiv subjects

Wenjie Gao

Publications and source records attributed to Wenjie Gao.

5 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↗

arXiv2024

Complementing Onboard Sensors with Satellite Map: A New Perspective for HD Map Construction

High-definition (HD) maps play a crucial role in autonomous driving systems. Recent methods have attempted to construct HD maps in real-time using vehicle onboard sensors. Due to the inherent limitations of onboard sensors, which include sensitivity to detection range and susceptibility to occlusion by nearby vehicles, the performance of these methods significantly declines in complex scenarios and long-range detection tasks. In this paper, we explore a new perspective that boosts HD map construction through the use of satellite maps to complement onboard sensors. We initially generate the satellite map tiles for each sample in nuScenes and release a complementary dataset for further research. To enable better integration of satellite maps with existing methods, we propose a hierarchical fusion module, which includes feature-level fusion and BEV-level fusion. The feature-level fusion, composed of a mask generator and a masked cross-attention mechanism, is used to refine the features from onboard sensors. The BEV-level fusion mitigates the coordinate differences between features obtained from onboard sensors and satellite maps through an alignment module. The experimental results on the augmented nuScenes showcase the seamless integration of our module into three existing HD map construction methods. The satellite maps and our proposed module notably enhance their performance in both HD map semantic segmentation and instance detection tasks.

cs.CV↗

arXiv2013

Extinction of solutions to a class of fast diffusion systems with nonlinear sources

In this paper, the finite time extinction of solutions to the fast diffusion system $u_t=\mathrm{div}(|\nabla u|^{p-2}\nabla u)+v^m$, $v_t=\mathrm{div}(|\nabla v|^{q-2}\nabla v)+u^n$ is investigated, where $1 0$ and $Ω\subset \mathbb{R}^N\ (N\geq1)$ is a bounded smooth domain. After establishing the local existence of weak solutions, the authors show that if $mn>(p-1)(q-1)$, then any solution vanishes in finite time provided that the initial data are ``comparable"; if $mn=(p-1)(q-1)$ and $Ω$ is suitably small, then the existence of extinction solutions for small initial data is proved by using the De Giorgi iteration process and comparison method. On the other hand, for $1<p=q<2$ and $mn<(p-1)^2$, the existence of at least one non-extinction solution for any positive smooth initial data is proved.

math.AP↗

arXiv2013

Study of Solutions for a quasilinear Elliptic Problem With negative exponents

The authors of this paper deal with the existence and regularities of weak solutions to the homogenous $\hbox{Dirichlet}$ boundary value problem for the equation $-\hbox{div}(|\nabla u|^{p-2}\nabla u)+|u|^{p-2}u=\frac{f(x)}{u^α}$. The authors apply the method of regularization and $\hbox{Leray-Schauder}$ fixed point theorem as well as a necessary compactness argument to prove the existence of solutions and then obtain some maximum norm estimates by constructing three suitable iterative sequences. Furthermore, we find that the critical exponent of $m$ in $\|f\|_{L^{m}(Ω)}$. That is, when $m$ lies in different intervals, the solutions of the problem mentioned belongs to different $\hbox{Sobolev}$ spaces. Besides, we prove that the solution of this problem is not in $W^{1,p}_{0}(Ω)$ when $α>2$, while the solution of this problem is in $W^{1,p}_{0}(Ω)$ when $1<α<2$.

math.AP↗

arXiv2013

Singular Phenomena of Solutions for Nonlinear Diffusion Equations involving $p(x)$-\hbox{Laplacian} Operator

The authors of this paper study singular phenomena(vanishing and blowing-up in finite time) of solutions to the homogeneous $\hbox{Dirichlet}$ boundary value problem of nonlinear diffusion equations involving $p(x)$-\hbox{Laplacian} operator and a nonlinear source. The authors discuss how the value of the variable exponent $p(x)$ and initial energy(data) affect the properties of solutions. At the same time, we obtain the critical extinction and blow-up exponents of solutions.

math.AP↗