Junhao Hu

SRIP: A Self-Refined Iterative Pipeline to Integrate Whole Body MRI-CT Registration and Pseudo-CT Generation

Junhao Hu, Hamed Yousefi, Tyler J. Fraum, John Crandall et autres

Motivation: Typical pipelines for MR synthesizing pseudo-CT (pCT) consist of two sequential steps: a MRI-CT registration model and a pCT generation model. pCT generation is particularly challenging in the body due to the difficulty in cross-modality image registration. Goal(s): To improve the …

Huawei Cloud Model-as-a-Service on the CloudMatrix384 SuperPod

Ao Xiao, Bangzheng He, Baoquan Zhang, Baoxing Huai et autres

Scaled-out MoE LLMs and scaled-up SuperPods create new systems challenges for production Model-as-a-Service (MaaS), requiring disaggregation, low-latency communication, and decentralized serving. This report presents xDeepServe, the production serving system behind Huawei Cloud's MaaS offering on CloudMatrix384, a 48-server SuperPod with 384 Ascend …

LLMigrate: Transforming "Lazy" Large Language Models into Efficient Source Code Migrators

Yuchen Liu, Junhao Hu, Yingdi Shan, Ge Li et autres

Rewriting C code in Rust provides stronger memory safety, yet migrating large codebases such as the 32-million-line Linux kernel remains challenging. While rule-based translators (e.g., C2Rust) provide accurate yet largely unsafe Rust programs, recent Large Language Model (LLM) approaches produce more idiomatic, …

MPIC: Position-Independent Multimodal Context Caching System for Efficient MLLM Serving

Shiju Zhao, Junhao Hu, Rongxiao Huang, Guihai Chen

The context caching technique is employed to accelerate the Multimodal Large Language Model (MLLM) inference by prevailing serving platforms currently. However, this approach merely reuses the Key-Value (KV) cache of the initial sequence of prompt, resulting in full KV cache recomputation even …

Directional Diffusion-Style Code Editing Pre-training

Qingyuan Liang, Zeyu Sun, Qihao Zhu, Junhao Hu et autres

Code pre-trained models have shown promising effectiveness in various software engineering tasks. Among these tasks, many tasks are related to software evolution and/or code editing. However, existing code pre-trained models often overlook the real-world code editing data and the evolutionary nature of …

An Infrared Image Detection of Substation Equipment integrating Selective Deblurring and Center2Net

Jinheng Li, Hanbo Zheng, Junhao Hu, Yanshen Liang et autres

The current Unmanned Aerial Vehicle (UAV) inspection scheme lacks infrared image preprocessing to deblur the image, and the detection robustness is not high in complex backgrounds. A method for detecting infrared images of substation equipment is proposed based on selective deblurring and …

cn (code pays fourni par la source)

Enhanced electromagnetic wave absorption of three-dimensional flower-like ZnO/TiO2/Ti3C2Tx composites

Yongpeng Wang, Shuyue Feng, Mengzhu Liu, Junhao Hu et autres

cn (code pays fourni par la source)

Angle-Insensitive Ultrathin Broadband Visible Absorber Based on Dielectric–Semiconductor–Lossy Metal Film Stacks

Yuanchen Ma, Junhao Hu, Wenfeng Li, Zhengmei Yang

Ultrathin broadband absorbers with high efficiency, wide angular tolerance, and low fabrication cost are in demand for various applications. Here, we present an angle-insensitive ultrathin (<150 nm) broadband absorber with an average 96.88% (experiment) absorptivity in the whole visible range by utilizing …

cn (code pays fourni par la source)

CupCleaner: A Hybrid Data Cleaning Approach for Comment Updating

Qingyuan Liang, Zeyu Sun, Qihao Zhu, Junhao Hu et autres

Comment updating is an emerging task in software evolution that aims to automatically revise source code comments in accordance with code changes. This task plays a vital role in maintaining code-comment consistency throughout software development. Recently, deep learning-based approaches have shown great …

3D flower-branch hierarchical carbon nanofiber/ZnO/TiO2/Ti3C2Tx composites for dramatically enhanced electromagnetic absorbing properties

Mengzhu Liu, Junhao Hu, Hongwei Wang, Yongpeng Wang

cn (code pays fourni par la source)

Marangoni Flow Manipulated Concentric Assembly of Cellulose Nanocrystals (Small Methods 11/2021)

Rongrong Shao, Xiao Meng, Zhaojie Shi, Jiajia Zhong et autres

Inside Back Cover In article number 2100690, Ye and co-workers report the concentric assembly of cellulose nanocrystals utilizing the Marangoni effect, indicating a uniform Maltese cross optical pattern under linearly polarized light. The assembly and optical property can be readily tuned to …

cn (code pays fourni par la source)

Marangoni Flow Manipulated Concentric Assembly of Cellulose Nanocrystals

Rongrong Shao, Xiao Meng, Zhaojie Shi, Jiajia Zhong et autres

Tunable assembly of cellulose nanocrystals (CNCs) is important for a variety of emerging applications in optics, sensing, and security. Most exploited assembly and optical property of CNCs are cholesteric assembly and corresponding circular dichroism. However, it still remains challenge to obtain homogenous …

cn (code pays fourni par la source)