A5100627166- Topic Modeling
- Natural Language Processing Techniques
- Software Engineering Research
- Software Engineering Techniques and Practices
- Advanced Computational Techniques and Applications
- Technology Assessment and Management
- Oceanographic and Atmospheric Processes
- Sentiment Analysis and Opinion Mining
- Simulation and Modeling Applications
- Video Surveillance and Tracking Methods
- Visual Attention and Saliency Detection
- Open Source Software Innovations
- Ocean Waves and Remote Sensing
- Web Data Mining and Analysis
- Phytochemistry and Biological Activities
- Network Security and Intrusion Detection
- Handwritten Text Recognition Techniques
- Geological Modeling and Analysis
- Computer Graphics and Visualization Techniques
- Meat and Animal Product Quality
- Gait Recognition and Analysis
- Urban and Freight Transport Logistics
- Digital literacy in education
- Occupational Health and Safety Research
- Human Pose and Action Recognition
Northeastern University
2024
Wuhan University
2021-2024
Dalian University of Technology
2008-2024
University of Hong Kong
2024
Xi’an Jiaotong-Liverpool University
2023
Dalian Naval Academy
2014-2023
Ocean University of China
2023
University of Ottawa
2023
University of California, San Diego
2023
Shanghai Business School
2022
Unsupervised person re-identification (re-ID) has attracted increasing research interests because of its scalability and possibility for real-world applications. State-of-the-art unsupervised re-ID methods usually follow a clustering-based strategy, which generates pseudo labels by clustering maintains memory to store instance features represent the centroid clusters contrastive learning. This approach suffers two problems. First, generated learning may not be perfect prototype. Forcing...
Large Language Models (LLMs) have revolutionized text generation, making detecting machine-generated increasingly challenging. Although past methods achieved good performance on pure text, those detectors poor distinguishing machine-revised (rewriting, expansion, and polishing), which can only minor changes from its original human prompt. As the content of may originate prompts, often involves identifying distinctive machine styles, e.g., worded favored by LLMs. However, existing struggle to...
Augmenting large language models (LLMs) with external tools has emerged as a promising approach to solving complex problems. However, traditional methods, which finetune LLMs tool demonstration data, can be both costly and restricted predefined set of tools. Recent in-context learning paradigm alleviates these issues, but the limited context length only allows for few shots demonstrations, leading suboptimal understandings Moreover, when there are numerous choose from, could completely fail...
Internal solitary waves (ISWs) in the South China Sea (SCS) are considerably modulated by background currents. In this study, a three-dimensional high-resolution non-hydrostatic model is configured to investigate how Kuroshio influences generation and evolution of ISWs northern SCS. Three runs conducted, including one control experiment without two sensitivity experiments with different paths. Luzon Strait (LS), reduces westward baroclinic energy flux radiated into SCS, resulting weakened...
Release planning for mobile apps has recently become an area of active research. Prior research in this concentrated on the analysis release notes and tracking user reviews to support app evolution with issue trackers. However, little is known about impact apps. Our work explores role updates based notes. For purpose, we collected Spotify, 'number one' 'Music' category Apple App Store, as data. Then, manually removed non-informative parts each note, determined relevance respect We did by...
Artificial Intelligence (AI) technologies have been developed rapidly, and AI-based systems widely used in various application domains with opportunities challenges. However, little is known about the architecture decisions made development, which has a substantial impact on success sustainability of these systems. To this end, we conducted an empirical study by collecting analyzing data from Stack Overflow (SO) GitHub. More specifically, searched SO six sets keywords explored 32 projects...
Large Language Models (LLMs) have greatly advanced code auto-completion systems, with a potential for substantial productivity enhancements developers. However, current benchmarks mainly focus on single-file tasks, leaving an assessment gap more complex, real-world, multi-file programming scenarios. To fill this gap, we introduce RepoBench, new benchmark specifically designed evaluating repository-level systems. RepoBench supports both Python and Java consists of three interconnected...
By using the ASEAN region as a case study, research attempts to examine how climate change has an impact on financial development, foreign direct investment, and economic growth. is considered among highest growing across world hence this study assesses consequences by taking into consideration growth of region. This conducted empirical test regression focuses five nations with greatest economies in These countries are Singapore, Malaysia, Thailand, Philippines, Indonesia, they cover years...
Most existing person re-identification (Re-ID) methods rely on high-cost manual annotations. To overcome the applicable issue, we focus a novel semi-supervised Re-ID without cross-camera annotations, which call random camera supervised (RCS). It is beneficial to real-world application, since short-time and cheap annotation conductive rapid deployment of re-ID. But only small proportion identities are annotated under camera, extremely challenging for pairing labeling or labeled image each...
Abstract Prefabricated construction has less environmental pollution, resource consumption, and high productivity. This new model is an important tool for the industry to achieve sustainable development. However, disruptions in prefabricated supply chain (PCSC) frequently occur practice, which seriously reduces performance of building projects. Improving resilience (RPCSC) urgent problem be solved. study first identified factors influencing RPCSC through a comprehensive literature review....
Semantic segmentation in bird's eye view (BEV) plays a crucial role autonomous driving. Previous methods usually follow an end-to-end pipeline, directly predicting the BEV map from monocular RGB inputs. However, challenge arises when inputs and targets distinct perspectives, making direct point-to-point hard to optimize. In this paper, we decompose original task into two stages, namely reconstruction RGB-BEV feature alignment. first stage, train autoencoder reconstruct maps given corrupted...
Generating accurate step-by-step reasoning is essential for Large Language Models (LLMs) to address complex problems and enhance robustness interpretability. Despite the flux of research on developing advanced approaches, systematically analyzing diverse LLMs strategies in generating chains remains a significant challenge. The difficulties stem from lack two key elements: (1) an automatic method evaluating generated different tasks, (2) unified formalism implementation approaches systematic...
In order to maximize the benefit of building supply chain, topology optimization method digital chain based on genetic neural network is studied. According overall structural characteristics customer demand data, supplier sales manufacturer production management and environmental policy exception data are collected form a set. As input prediction model improved LSTM constructed. Capture needs chain; according suppliers, manufacturers customers in nonlinear 0–1 mixed integer programming...
Large Language Models (LLMs) are reported to hold undesirable attestation bias on inference tasks: when asked predict if a premise P entails hypothesis H, instead of considering H's conditional truthfulness entailed by P, LLMs tend use the out-of-context truth label H as fragile proxy. In this paper, we propose pipeline that exploits do explicit inductive inference. Our uses an LLM transform into set attested alternatives, and then aggregate answers derived new entailment inquiries support...
Abstract. Innovation drives modern industry and enhances enterprise competitiveness. Despite China's progress, evident by its 14th position in the Global Index 2020, a gap remains compared to Western developed nations. Enterprises are key national innovation, especially rapidly evolving tech landscape. Big data presents new opportunities challenges for innovation. This study explores how enterprises can leverage big improve innovation outcomes, identifying factors that influence this...