A5101706260- Speech Recognition and Synthesis
- Natural Language Processing Techniques
- Topic Modeling
- Advanced Bandit Algorithms Research
- Internet Traffic Analysis and Secure E-voting
- Image and Video Quality Assessment
- Anomaly Detection Techniques and Applications
- Network Security and Intrusion Detection
- Recommender Systems and Techniques
- Advanced Decision-Making Techniques
Peng Cheng Laboratory
2023-2024
United States Naval Academy
2013
Duration bias widely exists in video recommendations, where models tend to recommend short videos for the higher ratio of finish playing and thus possibly fail capture users' true interests. In this paper, we eliminate duration from both data model. First, based on extensive analysis, observe that play completion rate with same presents a bimodal distribution. Hence, propose perform threshold division construct binary labels as training alleviating drawback overly biased towards videos....
Finetuning large language models (LLMs) has been empirically effective on a variety of downstream tasks. Existing approaches to finetuning an LLM either focus parameter-efficient finetuning, which only updates small number trainable parameters, or attempt reduce the memory footprint during training phase finetuning. Typically, stems from three contributors: model weights, optimizer states, and intermediate activations. However, existing works still require considerable none can...