Practice Quiz

According to Amazon Bedrock documentation, models accessed via API calls do not recall prior prompts and previous requests unless the preceding conversational context is explicitly included within the current prompt.

To reduce hallucinations, Amazon Bedrock guidelines recommend refining prompts using prompt optimization techniques, using techniques like Retrieval Augmented Generation (RAG) to give the model access to relevant data, or switching to a different model that may produce improved results.

According to Anthropic Claude guidelines, it is a good practice to use <example></example> tags for demonstration examples and to use different delimiters like H: and A: inside those examples to prevent confusion with the overall prompt's Human: and Assistant: delimiters.

According to Amazon Bedrock prompt engineering guidelines, a single prompt can combine multiple components such as the task or instruction, the context of the task, demonstration examples, and the input text.

Amazon Bedrock models accessed via API do not remember previous requests and prompts unless the conversation history is explicitly included in the current prompt. Therefore, to build a stateful chat application, developers must manually persist and pass the conversational history in each subsequent API request. Other options are incorrect as Bedrock does not automatically store session cache for APIs.

When prompting Anthropic Claude models with few-shot examples, it is best practice to use <example></example> tags for examples and delimiters like H: and A: to avoid conflict with Human: and Assistant: tags. Other options are incorrect as the User/Bot syntax is for Titan models, and using Human/Assistant delimiters within examples can cause confusion.

Amazon SageMaker Lakehouse is the capability that unifies data access across Amazon S3 data lakes, Amazon Redshift, and other data sources. This allows data engineers to streamline workflows without needing to build custom connections or copy data.

Amazon Bedrock is a service designed to enable users to build and scale generative AI applications. While Amazon SageMaker AI can build, train, and deploy models, Bedrock is focused specifically on scaling generative AI.

Amazon SageMaker Data and AI Governance allows users to discover, govern, and collaborate on data and AI securely with Amazon SageMaker Catalog, built on Amazon DataZone.

To reduce hallucinations, one can refine the prompt using optimization techniques, implement Retrieval Augmented Generation (RAG) to provide relevant data, or utilize a different model entirely.

Amazon Bedrock API calls are stateless and do not recall previous prompts or interactions unless the developer explicitly passes that interaction context in the current prompt payload.

Anthropic Claude models require specific conversational tags, specifically \\n\\nHuman: and \\n\\nAssistant:, within prompts sent via the Amazon Bedrock API, unlike Titan or Jurassic models.

When accessing models via the Amazon Bedrock API, the interaction is stateless. Models do not recall prior prompts or previous requests. To maintain a conversational-style interaction, the developer must manually include the previous interactions within the current prompt. Additionally, prompts sent to Anthropic Claude models via the API must be wrapped in conversational-style delimiters including '\\n\\nHuman:' and '\\n\\nAssistant:', whereas Titan Text and AI21 Labs Jurassic models do not require this wrapper, but still lack memory.

According to Amazon Bedrock documentation, there are three primary ways to reduce hallucinations in large language models: refining prompts using prompt optimization techniques, using techniques like Retrieval Augmented Generation (RAG) to provide the model with relevant data, or selecting and using a different model that may yield better performance.

For Anthropic Claude models, it is a recommended practice to wrap demonstration examples in <example></example> XML tags. Additionally, to prevent the model from confusing the examples with the main conversational flow, developers should use alternative delimiters like H: and A: inside those examples instead of the global prompt delimiters Human: and Assistant:.

Amazon SageMaker Unified Studio is a single development environment that enables users to build using all their tools and data for analytics and AI, making it the perfect fit for the business constraint.

Amazon SageMaker Data and AI Governance enables secure discovery, governance, and collaboration on data and AI using the Amazon SageMaker Catalog, which is built on Amazon DataZone.

According to the official definition, Amazon SageMaker AI is a fully managed machine learning service that allows developers and data scientists to build, train, and deploy machine learning and foundation models.