AI Tools & Technologies Flashcards

Which technique adapts a pre-trained LLM to a specific task by training only a small set of added parameters, leaving the original weights frozen?

Answer: LoRA (Low-Rank Adaptation)

LoRA inserts low-rank weight matrices into transformer layers and trains only those, making fine-tuning far cheaper than updating all model parameters.

A consultant is evaluating an AI pipeline for hallucinations. Which evaluation approach uses another LLM to judge response quality and factual grounding?

Answer: LLM-as-a-judge evaluation

LLM-as-a-judge uses a capable model (e.g., GPT-4) to automatically evaluate responses for accuracy, relevance, and grounding.

Which multimodal AI model from OpenAI can process both images and text as input to answer questions about visual content?

Answer: GPT-4V (GPT-4 with Vision)

GPT-4V extends GPT-4 with vision capabilities, enabling it to analyze and answer questions about images alongside text.

What is the main advantage of using streaming responses in LLM API calls?

Answer: It delivers tokens to the user incrementally as they are generated, improving perceived latency

Streaming sends tokens to the client as they are generated rather than waiting for the full response, making the application feel faster to users.

Which open-source framework developed by Microsoft is used to build multi-agent AI systems where agents can communicate and collaborate?

Answer: AutoGen

Microsoft's AutoGen is an open-source framework for building multi-agent systems where LLM-powered agents converse and collaborate to solve tasks.

When chunking documents for a RAG system, what is the primary trade-off between smaller and larger chunk sizes?

Answer: Smaller chunks improve retrieval precision but may lose context; larger chunks preserve context but may dilute relevance

Smaller chunks retrieve more targeted passages but may miss surrounding context, while larger chunks retain context but can introduce irrelevant content into retrieved results.

Which AI model deployment pattern uses a lightweight 'router' model to direct queries to the most cost-effective or capable model based on complexity?

Answer: LLM routing

LLM routing dynamically sends simple queries to cheaper, faster models and complex ones to more capable models, optimizing cost and latency.