Knowledge Check

The following questions will help you check what you've learned.

1. Question 1 of 4Which option correctly describes the concept of negative prompts in prompt engineering?

   ANegative prompts are additional instructions provided to the language model to guide it towards generating undesired outputs.BNegative prompts are examples of incorrect or undesirable outputs that the language model should avoid generating.CNegative prompts are context or background information that the language model should ignore while generating responses.DNegative prompts are techniques used to reduce the dimensionality of the language model's latent space.
   Check answer
2. Question 2 of 4A data scientist at AnyCompany wants to influence a model's responses by adjusting various inference parameters to achieve the desired level of diversity, coherence, and relevance in the generated outputs. Which statement correctly describes the effect of increasing the temperature parameter during inference?

   AIncreasing the temperature parameter will make the model's responses more deterministic and less diverse, favoring the most probable outputs.BIncreasing the temperature parameter will make the model's responses more diverse and less deterministic, exploring a wider range of possible outputs.CIncreasing the temperature parameter will have no effect on the diversity of the model's responses, because it only impacts the length of the generated outputs.DIncreasing the temperature parameter will make the model's responses shorter and more concise, because it reduces the probability of generating longer sequences.
   Check answer
3. Question 3 of 4AnyCompany is developing a generative artificial intelligence (generative AI) system to assist customers in filing their tax returns. The process of filing taxes involves several complex steps, such as gathering relevant documents, calculating income and deductions, and completing various forms and schedules. Which approach to prompt engineering would be most effective for this tax filing assistance system?

   AUse a single, comprehensive prompt that covers the entire tax filing process.BProvide the language model with a detailed prompt that explains the tax filing process from start to finish.CRely solely on the language model's pretraining on tax-related data, without any additional prompt engineering.DBreak down the tax filing process into smaller, more manageable tasks, and craft separate prompts for each task.
   Check answer
4. Question 4 of 4An organization is using a generative model to solve complex mathematical word problems. The prompt used must break down the problem, identify relevant information, and apply appropriate problem-solving strategies and calculations. Which prompt engineering technique would be most effective for this mathematical problem-solving?

   AZero-shot prompting: Provide the language model with a single prompt containing the word problem and ask it to solve it.BFew-shot prompting: Include a small number of examples of solved word problems in the prompt.CChain-of-thought prompting: Prompt the language model to break down the problem-solving process into a series of logical steps and show its step-by-step reasoning.DChain-of-thought prompting combined with few-shot prompting: Prompt the model to break down the problem-solving process into a series of logical steps, and include a few examples of solved word problems.
   Check answer

Navigate to the next lesson to learn about some of the misuses and risks involved in prompt engineering.