Independently-Normalized SGD for Generalized-Smooth Nonconvex Optimization

Kavli Affiliate: Yi Zhou | First 5 Authors: Yufeng Yang, Erin Tripp, Yifan Sun, Shaofeng Zou, Yi Zhou | Summary: Recent studies have shown that many nonconvex machine learning problems meet a so-called generalized-smooth condition that extends beyond traditional smooth nonconvex optimization. However, the existing algorithms designed for generalized-smooth nonconvex optimization encounter significant limitations in […]


Continue.. Independently-Normalized SGD for Generalized-Smooth Nonconvex Optimization

Physical Space Proof of Bilinear Estimates and Applications to Nonlinear Dispersive Equations

Kavli Affiliate: Yi Zhou | First 5 Authors: Li Tu, Yi Zhou, , , | Summary: We give a simpler proof for the local well-posedness of the modified Korteweg-de Vries equations and modified Benjamin-Ono equation in $H^{frac{1}{4}}(mathbb{R})$ and $H^{frac{1}{2}}(mathbb{R})$, respectively. The proof is based on the Strichartz estimate, dyadic decomposition and a bilinear estimate given […]


Continue.. Physical Space Proof of Bilinear Estimates and Applications to Nonlinear Dispersive Equations

DriveDreamer4D: World Models Are Effective Data Machines for 4D Driving Scene Representation

Kavli Affiliate: Zheng Zhu | First 5 Authors: Guosheng Zhao, Chaojun Ni, Xiaofeng Wang, Zheng Zhu, Guan Huang | Summary: Closed-loop simulation is essential for advancing end-to-end autonomous driving systems. Contemporary sensor simulation methods, such as NeRF and 3DGS, rely predominantly on conditions closely aligned with training data distributions, which are largely confined to forward-driving […]


Continue.. DriveDreamer4D: World Models Are Effective Data Machines for 4D Driving Scene Representation

DriveDreamer4D: World Models Are Effective Data Machines for 4D Driving Scene Representation

Kavli Affiliate: Zheng Zhu | First 5 Authors: Guosheng Zhao, Chaojun Ni, Xiaofeng Wang, Zheng Zhu, Xueyang Zhang | Summary: Closed-loop simulation is essential for advancing end-to-end autonomous driving systems. Contemporary sensor simulation methods, such as NeRF and 3DGS, rely predominantly on conditions closely aligned with training data distributions, which are largely confined to forward-driving […]


Continue.. DriveDreamer4D: World Models Are Effective Data Machines for 4D Driving Scene Representation

DriveDreamer4D: World Models Are Effective Data Machines for 4D Driving Scene Representation

Kavli Affiliate: Zheng Zhu | First 5 Authors: Guosheng Zhao, Chaojun Ni, Xiaofeng Wang, Zheng Zhu, Xueyang Zhang | Summary: Closed-loop simulation is essential for advancing end-to-end autonomous driving systems. Contemporary sensor simulation methods, such as NeRF and 3DGS, rely predominantly on conditions closely aligned with training data distributions, which are largely confined to forward-driving […]


Continue.. DriveDreamer4D: World Models Are Effective Data Machines for 4D Driving Scene Representation

Measuring Non-Hermitian Topological Invariants Directly from Quench Dynamics

Kavli Affiliate: Long Zhang | First 5 Authors: Xiao-Dong Lin, Long Zhang, , , | Summary: While non-Hermitian (NH) topological phases and phenomena have been observed across various quantum systems, directly measuring NH topological invariants remains a significant challenge. In this study, we present a generic and unified framework for the direct measurement of various […]


Continue.. Measuring Non-Hermitian Topological Invariants Directly from Quench Dynamics

DreamCraft3D++: Efficient Hierarchical 3D Generation with Multi-Plane Reconstruction Model

Kavli Affiliate: Cheng Peng | First 5 Authors: Jingxiang Sun, Cheng Peng, Ruizhi Shao, Yuan-Chen Guo, Xiaochen Zhao | Summary: We introduce DreamCraft3D++, an extension of DreamCraft3D that enables efficient high-quality generation of complex 3D assets. DreamCraft3D++ inherits the multi-stage generation process of DreamCraft3D, but replaces the time-consuming geometry sculpting optimization with a feed-forward multi-plane […]


Continue.. DreamCraft3D++: Efficient Hierarchical 3D Generation with Multi-Plane Reconstruction Model

WorldCuisines: A Massive-Scale Benchmark for Multilingual and Multicultural Visual Question Answering on Global Cuisines

Kavli Affiliate: Yi Zhou | First 5 Authors: Genta Indra Winata, Frederikus Hudi, Patrick Amadeus Irawan, David Anugraha, Rifki Afina Putri | Summary: Vision Language Models (VLMs) often struggle with culture-specific knowledge, particularly in languages other than English and in underrepresented cultural contexts. To evaluate their understanding of such knowledge, we introduce WorldCuisines, a massive-scale […]


Continue.. WorldCuisines: A Massive-Scale Benchmark for Multilingual and Multicultural Visual Question Answering on Global Cuisines

WorldCuisines: A Massive-Scale Benchmark for Multilingual and Multicultural Visual Question Answering on Global Cuisines

Kavli Affiliate: Yi Zhou | First 5 Authors: Genta Indra Winata, Frederikus Hudi, Patrick Amadeus Irawan, David Anugraha, Rifki Afina Putri | Summary: Vision Language Models (VLMs) often struggle with culture-specific knowledge, particularly in languages other than English and in underrepresented cultural contexts. To evaluate their understanding of such knowledge, we introduce WorldCuisines, a massive-scale […]


Continue.. WorldCuisines: A Massive-Scale Benchmark for Multilingual and Multicultural Visual Question Answering on Global Cuisines