C1000-185問題集、IBM実際の試験問題

質問 1

In the context of Retrieval-Augmented Generation (RAG) models, embeddings play a crucial role in retrieving relevant documents.
Which of the following best describes the purpose of embeddings in GenAI, particularly within a RAG system?

A. Embeddings are primarily used to measure the grammatical correctness of the generated responses by checking the sentence structure in the generated text.

B. Embeddings are dense vector representations of words, phrases, or entire documents that capture semantic relationships, making it easier to find similar content in the knowledge base during retrieval.

C. Embeddings are used to compress large documents into a smaller format, enabling faster storage and retrieval within a RAG system.

D. Embeddings represent each word in a fixed-length vector based on its semantic meaning, which allows for precise retrieval of documents by directly matching words between the query and the documents.

正解: B

質問 2

In a Retrieval-Augmented Generation (RAG) system designed for technical document retrieval, you are tasked with implementing text chunking techniques using the LangChain library. The technical documents are large and contain numerous tables, figures, and bullet points.
What is the most effective way to handle text splitting to ensure high-quality retrieval?

A. Split the text into equal-sized chunks of 512 characters, regardless of the content structure, to improve consistency in retrieval.

B. Use a hybrid approach, splitting the text by both semantic boundaries (like paragraphs) and content-specific markers (like bullet points and tables), while keeping chunks within the model's token limit.

C. Split the text only at paragraph breaks, ignoring tables and figures, as they can be processed separately.

D. Convert tables and figures into plain text and split the document by character count to maintain even chunk sizes.

正解: B

質問 3

You are part of a team building an AI-powered assistant that helps software developers by answering technical queries. To handle the vast amount of technical documentation efficiently, the team has chosen to implement the Retrieval-Augmented Generation (RAG) pattern. LangChain is used to construct the retrieval-generation pipeline, and SingleStore is employed to store the document embeddings. You need to ensure that the integration of LangChain with SingleStore allows for real-time retrieval of relevant technical documentation.
Which of the following configurations should be used to ensure an optimal RAG implementation with LangChain and SingleStore?

A. Use LangChain to retrieve pre-generated embeddings from SingleStore, and then fine-tune the generative model based on the retrieved embeddings to improve response accuracy.

B. Use LangChain to embed documents into a vector space and store these embeddings in SingleStore. During inference, embed the query in the same vector space and retrieve documents using vector similarity search.

C. Use LangChain to retrieve previously generated embeddings from SingleStore, then match these embeddings against the query using cosine similarity before passing them to the generative model.

D. Use LangChain to embed both queries and documents into the same vector space and store these embeddings in SingleStore. During inference, retrieve documents using SQL queries based on keyword similarity.

正解: B

質問 4

A financial services company is building a generative AI model to assist with customer support. The company is concerned about potential legal liabilities if the model generates customer information, such as bank account numbers or personal identification data, as part of its responses.
Which of the following techniques would best mitigate the risk of generating Personally Identifiable Information (PII) during inference?

A. Set a strict token limit to prevent the model from generating long sequences, assuming PII tends to appear in longer outputs.

B. Use greedy decoding to ensure the model generates only the most probable tokens, which are less likely to include PII.

C. Train the model on sensitive customer data but ensure that the temperature is set low to avoid generating diverse outputs.

D. Implement a real-time PII filter that detects and removes sensitive data before the output is presented to the user.

正解: D

質問 5

In the context of a Retrieval-Augmented Generation (RAG) system, which type of retriever is best suited for retrieving documents based on semantic similarity in a vector space?

A. Dense retriever, which converts both the query and documents into vectors and retrieves based on similarity in the vector space.

B. Hierarchical retriever, which first applies keyword-based filters and then ranks documents using machine learning models.

C. Boolean retriever, which uses logical operators (AND, OR, NOT) to filter documents based on keyword presence.

D. Exact match retriever, which retrieves documents that exactly match the query string.

正解: A

質問 6

In the context of quantizing large language models (LLMs), which of the following statements best describes the key trade-offs between model size, performance, and accuracy when using quantization techniques?

A. Quantization always improves model performance but significantly increases model size.

B. Quantization maintains model accuracy but doubles the computation required for inference.

C. Quantization eliminates the need for fine-tuning after deployment, ensuring zero accuracy loss.

D. Quantization reduces model size but may lead to a loss of accuracy, especially with aggressive quantization methods.

正解: D

質問 7

In the context of using Tuning Studio in IBM watsonx, which of the following is the primary benefit of fine-tuning a pre-trained Generative AI model using this tool?

A. Increases the size of the model to handle more complex prompts.

B. Allows you to train the model from scratch on a new dataset.

C. Reduces the time needed to generate text responses by simplifying the model's architecture.

D. Improves model performance on domain-specific tasks without retraining from scratch.

正解: D

質問 8

You are tasked with configuring a generative model to assist legal professionals by generating contract clauses. The model should produce complete, contextually relevant clauses but should avoid overly long and redundant text. You must ensure that the generation process stops appropriately once a clause is completed.
Which of the following strategies is the most appropriate for configuring stopping criteria?

A. Set a maximum token limit of 500 and implement a fixed length stopping criterion.

B. Set the top-k value to 1 and use the default maximum token limit, relying on the model's internal stopping criteria.

C. Use a beam search algorithm and impose a strict length restriction of 100 tokens.

D. Use a stop sequence based on legal clause delimiters and impose a minimum token limit of 50.

正解: D

IBM watsonx Generative AI Engineer - Associate - C1000-185 模擬練習