A5094239331- Recommender Systems and Techniques
- Embedded Systems Design Techniques
- Machine Learning and Algorithms
- Parallel Computing and Optimization Techniques
- Real-Time Systems Scheduling
- Speech and dialogue systems
- AI-based Problem Solving and Planning
- Machine Learning and Data Classification
- Machine Learning and ELM
- Handwritten Text Recognition Techniques
City University of Hong Kong
2024-2025
To enhance the efficacy of multi-scenario services in industrial recommendation systems, emergence multi-domain has become prominent, which entails simultaneous modeling all domains through a unified model, effectively capturing commonalities and differences among them. However, current methods rely on manual domain partitioning, overlook intricate relationships heterogeneity different during joint optimization, hindering integration differences. address these challenges, this paper proposes...
Deep Recommender Systems (DRS) are increasingly dependent on a large number of feature fields for more precise recommendations. Effective selection methods consequently becoming critical further enhancing the accuracy and optimizing storage efficiencies to align with deployment demands. This research area, particularly in context DRS, is nascent faces three core challenges. Firstly, variant experimental setups across papers often yield unfair comparisons, obscuring practical insights....
The reranker and generator are two critical components in the Retrieval-Augmented Generation (i.e., RAG) pipeline, responsible for ranking relevant documents generating responses. However, due to differences pre-training data objectives, there is an inevitable gap between ranked as by those required support answering query. To address this gap, we propose RADIO, a novel practical preference alignment framework with RAtionale DIstillatiOn. Specifically, We first rationale extraction method...
The performance of Dense retrieval (DR) is significantly influenced by the quality negative sampling. Traditional DR methods primarily depend on naive sampling techniques or mining hard negatives through external retriever and meticulously crafted strategies. However, often fails to adequately capture accurate boundaries between positive samples, whereas existing are prone false negatives, resulting in degradation training instability. Recent advancements large language models (LLMs) offer...