A financial institution uses Snowflake Cortex LLM functions to process customer feedback. They initially used SNOWF LAKE .CORTEX.SENTIMENT for general sentiment analysis. Now, they need to extract specific sentiment categories (e.g., 'service_quality', 'product_pricing') and the sentiment for each, expecting the output in a structured JSON format for automated downstream processing. Which AI_COMPLETE configuration best addresses their new requirement while considering cost-efficiency and output reliability?

• A.
  
• B.
  
• C.
  
• D.
  
• E.
  

Explanation: Only visible for TrainingDump members. You can sign-up / login (it's free).

A software development team is building a conversational AI application within Snowflake, aiming to provide a dynamic and stateful chat experience for users. The application needs to handle follow-up questions while maintaining context, provide responses with a degree of creative variation, and actively filter out any potentially harmful content. The team utilizes the SNOWFLAKE. CORTEX. COMPLETE (or AI_COMPLETE) function.

• A.
  
• B.
  
• C.
  
• D.
  
• E.
  

Explanation: Only visible for TrainingDump members. You can sign-up / login (it's free).

A company is developing a RAG application to provide concise and highly relevant answers to user queries from a vast knowledge base of technical documents. They are using Cortex Search for retrieval and are considering different embedding models and text chunking strategies to optimise the system. Which of the following statements about Cortex Search embedding models and RAG best practices are correct? (Select all that apply)

• A. The cost for embedding models in Cortex Search, such as 'snowflake-arctic-embed-l-v2.0' and 'e5-base-v2, is incurred based on both input and output tokens.
• B. The 'voyage-multilingual-2 model is suitable for multilingual documents and has a significantly larger context window (32000 tokens) compared to 'snowflake- arctic-embed-l-v2.C (512 tokens), making it more robust for longer text inputs.
• C. The
  
• D. Using the 'snowflake-arctic-embed-l-v2.0-8k' model, which has an 8192-token context window, allows processing entire large technical documents as a single chunk for embedding, leading to better RAG results by preserving full document context.
• E. For optimal RAG retrieval quality with Cortex Search, it is recommended to split text into chunks of no more than 512 tokens, even when using models with larger context windows like 'snowflake-arctic-embed-l-v2.0-8k'.

Explanation: Only visible for TrainingDump members. You can sign-up / login (it's free).

A data scientist needs to fine-tune a 'mistral-7b' LLM using Snowflake Cortex for a specific text summarization task. They have prepared a training dataset in a Snowflake table, with text in a column named 'source_text' and the desired summaries in a column named 'expected_summary' . They also want to understand the cost implications. Which SQL statement will correctly initiate the fine-tuning job, and how will the cost be primarily calculated?

• A. The fine-tuning job is initiated by:
  
• B. The fine-tuning job is initiated by:
  
• C. The fine-tuning job is initiated by providing the prompt and completion data directly as arrays within the 'FINETUNE' function call, avoiding the need for a separate training data table, and costs are only for the storage of the fine-tuned model.
• D. The fine-tuning job is initiated by:
  
• E. The fine-tuning job is initiated by:
  

Explanation: Only visible for TrainingDump members. You can sign-up / login (it's free).

An enterprise is deploying a Cortex Analyst application and needs to manage its cost, ensure data security, and understand its operational behavior within Snowflake. Which of the following statements are true regarding the deployment, cost, and security of Cortex Analyst and its semantic models?

• A. Semantic models for Cortex Analyst, whether stored as YAML files or native semantic views, should have their access controlled by RBAC. This implicitly controls access to the underlying tables referenced in the semantic model.
• B. The primary cost incurred for Cortex Analyst is based on the number of tokens processed by the underlying LLMs, with more complex natural language questions directly leading to higher token usage and charges.
• C. When using Snowflake-hosted LLMs (e.g., from Mistral or Meta) with Cortex Analyst, all customer data, including metadata and prompts, remains within Snowflake's governance boundary.
• D. Snowflake strongly recommends enabling the ENABLE_CORTEX_ANALYST_MODEL_AZURE_OPENAI account parameter to leverage Azure OpenAI models for Cortex Analyst, as it offers the highest performance and respects RBAC restrictions for these models.
• E. Administrators can monitor Cortex Analyst requests, including the user, question asked, generated SQL, and errors, by querying the SNOWFLAKLOCAL .CORTEX_ANALYST_REQUESTS function.

Explanation: Only visible for TrainingDump members. You can sign-up / login (it's free).