1. 多任务学习
如何提高multi-task的performance
2. 少标签分类
如何处理labelled data较少的情况
3. Design an AI system using Large Language Models (LLMs).
In an AI role, you may be tasked with designing an AI system using Large Language Models (LLMs). You should consider the limitations of LLMs, such as their need for large amounts of data and their tendency to generate plausible but incorrect or nonsensical answers. Your design should address how to integrate LLMs effectively, possibly with other models or modalities, to apply them to industry-specific projects.
4. Explain the Transformer architecture and its alignment mechanism.
During a machine learning interview, you might be asked to explain the Transformer architecture and its alignment mechanism. The Transformer is a model architecture that eschews recurrence and instead relies entirely on an attention mechanism to draw global dependencies between input and output. The alignment mechanism refers to the way the model aligns different parts of the input sequence with the output sequence, which is crucial for tasks like translation where the order of words can vary significantly between languages.
