A5101398010- Recommender Systems and Techniques
- Cryospheric studies and observations
- Arctic and Antarctic ice dynamics
- Advanced Graph Neural Networks
- Climate change and permafrost
- Multimodal Machine Learning Applications
- Mental Health via Writing
- Data Stream Mining Techniques
- Soil Moisture and Remote Sensing
- Scientific Computing and Data Management
- Image Retrieval and Classification Techniques
- Advanced Bandit Algorithms Research
- Advanced Image and Video Retrieval Techniques
Shenzhen University
2023-2024
University of Waterloo
2020-2023
Communitech
2020
Accurate cloud detection is critical for quantitative applications of satellite-based advanced imager observations, yet nighttime presents challenges due to the lack visible and near-infrared spectral information. Nighttime using infrared (IR)-only information needs be improved. Based on a collocated dataset from Fengyun-3D Medium Resolution Spectral Imager (FY-3D MERSI) Level 1 data CALIPSO CALIOP lidar 2 product, this study proposes novel framework leveraging Light Gradient-Boosting...
Existing Large Vision-Language Models (LVLMs) can process inputs with context lengths up to 128k visual and text tokens, yet they struggle generate coherent outputs beyond 1,000 words. We find that the primary limitation is absence of long output examples during supervised fine-tuning (SFT). To tackle this issue, we introduce LongWriter-V-22k, a SFT dataset comprising 22,158 examples, each multiple input images, an instruction, corresponding ranging from 0 10,000 Moreover, achieve maintain...
This study introduces the first use of Global Navigation Satellite System Reflectometry (GNSS-R) for monitoring lake ice phenology. is demonstrated using Qinghai Lake, Tibetan Plateau, as a case study. Signal-to-Noise Ratio (SNR) values obtained from Cyclone GNSS (CYGNSS) constellation over four seasons (2018 to 2022) were used examine impact surface conditions on reflected signals during open water and cover seasons. A moving t-test algorithm was applied time-varying SNR allowing detection...
Although knowledge graph has shown their effectiveness in mitigating data sparsity many recommendation tasks, they remain underutilized context-aware recommender systems (CARS) with the specific challenges associated contextual features, i.e., feature and interaction sparsity. To bridge this gap, paper, we propose a novel pairwise intent embedding learning (PING) framework to efficiently integrate graphs into CARS. Specifically, our PING contains three modules: 1) construction module is used...
Different from the data sparsity that traditional recommendations suffer from, context-aware recommender systems (CARS) face specific challenges related to contextual features, i.e., feature and interaction sparsity. How knowledge graphs address these remains under-discussed. To bridge this gap, in paper, we first propose a novel pairwise intent graph (PIG) containing nodes of users, items, entities, enhanced intents integrate into CARS efficiently. Enhanced are generated through fusion...
In evaluating the long-context capabilities of large language models (LLMs), benchmarks such as "Needle-in-a-Haystack" (NIAH), Ruler, and Needlebench are commonly used. While these measure how well understand input sequences, they do not effectively gauge quality long-form text generation--a critical aspect for applications design proposals creative writing. To address this gap, we have introduced a new evaluation benchmark, LongGenBench, which tests models' ability to identify specific...