A5102008510- Topic Modeling
- Web Data Mining and Analysis
- Natural Language Processing Techniques
- Recommender Systems and Techniques
- Information Retrieval and Search Behavior
- Anomaly Detection Techniques and Applications
- Semantic Web and Ontologies
- Geological Modeling and Analysis
- Business Process Modeling and Analysis
- Human Pose and Action Recognition
- Advanced Computational Techniques and Applications
- Model-Driven Software Engineering Techniques
- Video Analysis and Summarization
- Hand Gesture Recognition Systems
- Text and Document Classification Technologies
- Advanced Image and Video Retrieval Techniques
- Advanced SAR Imaging Techniques
- Time Series Analysis and Forecasting
- Privacy-Preserving Technologies in Data
- Service-Oriented Architecture and Web Services
- Advanced Graph Neural Networks
- Handwritten Text Recognition Techniques
- Image Retrieval and Classification Techniques
- Video Surveillance and Tracking Methods
- Dental Radiography and Imaging
Amazon (United States)
2023-2025
Search
2024
Dalian Ocean University
2023
Harbin Institute of Technology
2005-2013
Process-supervised reward models serve as a fine-grained function that provides detailed step-wise feedback to model responses, facilitating effective selection of reasoning trajectories for complex tasks. Despite its advantages, evaluation on PRMs remains less explored, especially in the multimodal domain. To address this gap, paper first benchmarks current vision large language (VLLMs) two types models: output (ORMs) and process (PRMs) multiple vision-language benchmarks, which reveal...
Graph Neural Networks (GNNs) have achieved great success in modeling graph-structured data. However, recent works show that GNNs are vulnerable to adversarial attacks which can fool the GNN model make desired predictions of attacker. In addition, training data be leaked under membership inference attacks. This largely hinders adoption high-stake domains such as e-commerce, finance and bioinformatics. Though investigations been made conducting robust protecting privacy, they generally fail...
Developing a universal model that can effectively harness heterogeneous resources and respond to wide range of personalized needs has been longstanding community aspiration. Our daily choices, especially in domains like fashion retail, are substantially shaped by multi-modal data, such as pictures textual descriptions. These modalities not only offer intuitive guidance but also cater user preferences. However, the predominant personalization approaches mainly focus on ID or text-based...
Large language models (LLM) have demonstrated emergent abilities in text generation, question answering, and reasoning, facilitating various tasks domains. Despite their proficiency tasks, LLMs like LaPM 540B Llama-3.1 405B face limitations due to large parameter sizes computational demands, often requiring cloud API use which raises privacy concerns, limits real-time applications on edge devices, increases fine-tuning costs. Additionally, underperform specialized domains such as healthcare...
Aiming at a better understanding of the search goals in user sessions, recent query recommender systems explicitly model reformulations queries, which hopes to estimate intents behind these and thus benefit next-query recommendation. However, real-world e-commercial scenarios, are much more complicated may evolve dynamically. Existing methods merely consider trivial reformulation from semantic aspects fail dynamic intent flows leading sub-optimal capacities recommend desired queries. To deal...
Foundation models such as GPT-4 for natural language processing (NLP), Flamingo computer vision (CV), have set new benchmarks in AI by delivering state-of-the-art results across various tasks with minimal task-specific data. Despite their success, the application of these to graph domain is challenging due relational nature graph-structured To address this gap, we propose Graph Model (GFM) Workshop, first workshop GFMs, dedicated exploring adaptation and development foundation specifically...
Online shopping platforms, such as Amazon, offer services to billions of people worldwide. Unlike web search or other engines, product engines have their unique characteristics, primarily featuring short queries which are mostly a combination attributes and structured space. The uniqueness underscores the crucial importance query understanding component. However, there limited studies focusing on exploring this impact within real-world engines. In work, we aim bridge gap by conducting...
This paper aims to present a long-term background memory framework, which is capable of memorizing long period in video and rapidly adapting the changes background. Based on Gaussian mixture model (GMM), this framework enables an accurate identification appearances presents perfect solution numerous typical problems foreground detection. The experimental results with various benchmark sequences quantitatively qualitatively demonstrate that proposed algorithm outperforms many GMM-based...
Semantic identifier (ID) is an important concept in information retrieval that aims to preserve the semantics of objects such as documents and items inside their IDs. Previous studies typically adopt a two-stage pipeline learn semantic IDs by first procuring embeddings using off-the-shelf text encoders then deriving based on embeddings. However, each step introduces potential loss there usually inherent mismatch between distribution within latent space produced anticipated required for...
With the increasing demand for sea treasures such as cucumbers, urchins, and shells in market, global aquaculture industry has been booming. However, traditional counting methods these have high labor costs, low efficiency, large errors. This article proposes an object detection optimization algorithm based on YOLOv7 network, which reduces redundant calculations memory access time through pruning quantization to improve efficiency of extracting spatial features. The is deployed Jetson Nano...
Focusing on the problem of low computation efficiency in process tracking human 3D motion, fast algorithm for arm motion based Joint-Chain Motion Model (JCMM) is proposed Particle Filter.In our algorithm, via defined, state space can be discomposed into some dimension subspaces, and amount particle reduced.The result experiment shows that advance computational while guarantee precision tracking.