In the data processing phase, what is a typical format for semi-structured data?

If you want to train a model to predict next month's house prices, which machine learning method should you prioritize?

What is the core difference in objectives between Discriminative AI (DAI) and Generative AI (GAI)?

In the Responsible AI framework, what are the two main principles that Explainable AI (XAI) aims to achieve?

Which statistical method should a bank use to quantify the deviation of large expenditure records in its financial report to determine if they are outliers?

In regression model evaluation, which error metric is most sensitive to extreme outliers?

In regression model evaluation, which error metric is most sensitive to extreme outliers?

According to the EU AI Act, if a company uses voice analysis to determine whether an employee is suitable for promotion, which risk level should this AI system be classified as?

In the ETL process of data warehousing, at which stage should inconsistent customer names from different systems be standardized?

Compared to RNNs, what is the core design in Transformer models that overcomes the long sequence memory bottleneck?

In defect detection in smart factories, if the cost of false negatives is high, which evaluation metric should be prioritized?

In reinforcement learning processes, what is the strategy algorithm used to align large language models (LLMs) with human preferences?

What two models make up a Generative Adversarial Network (GAN), and through what mechanism does it enhance the realism of generated images?

In an unordered list of 50 student scores, what is the most suitable search method to find a specific student's score?

When a set of house sale price data contains a few extreme high prices from luxury homes, which measure of central tendency best reflects the general price level?

Which central trend measurement indicator best reflects the general housing price level?

Compared to Low Code platforms, what is the biggest advantage and core limitation of No Code platforms in enterprise applications?

To avoid large language models (LLM) producing fictional or factually incorrect content (AI hallucination), which architecture should be prioritized?

The core principle of Zero Trust Architecture (ZTA) is to never trust and always verify. What risk does this cybersecurity mindset primarily address?

In adjusting output parameters of LLMs, how should the Temperature parameter be set to maximize creativity and diversity in generated text?

Comparing RAG and Fine-tuning, which option correctly presents their advantages and disadvantages in terms of knowledge timeliness and style alignment?

In manufacturing, if real-time monitoring of machine data and issuing alerts during anomalies is required, which inference model and deployment strategy should be adopted?

What is the core purpose of the Proof of Concept (PoC) phase in the AI implementation process?

To comply with regulations such as GDPR, what is the technique called when companies add random noise to customer data before training LLMs to obscure individual data points?

When evaluating Low Code platforms, if the API interfaces provided by the platform are insufficient, what two risks will increase?

What is the concept that protects individual privacy by adding noise to data to prevent sensitive information from being inferred?

When evaluating a Low Code platform, what two risks are heightened if the API interfaces provided by the platform are insufficient?

What module is primarily relied upon to transform the AI role from passive decision support to active task executor?

What is the benchmark dataset promoted by Taiwan's Digital Development Department to evaluate local LLM performance in Taiwan-specific knowledge?

If an LLM is required to act as a legal advisor for complex contract logical analysis, which prompt strategy and parameter setting should be adopted to ensure accuracy?

If a designer uses AI to generate an image of 4096 x 4096 pixels, what is the approximate size in inches if printed at 600 DPI?

What is the communication standard used to control, transmit, and synchronize messages between agents in complex workflows involving multiple AI agents?

In the implementation/operation phase of AI, what is the correct order of steps to continuously leverage AI value and address model performance decline?