AIF-C01 Question 145
Select 3You are tasked with fine-tuning a foundation model for a customer service chatbot. The goal is to ensure the model provides accurate responses and aligns with company policies. Which actions should you take to prepare the data for fine-tuning?
- A
Ensure the dataset includes representative examples of common customer queries and responses.
- B
Randomly sample unstructured text data from the internet without filtering.
- C
Apply data labeling to classify customer queries based on intent and sentiment.
- D
Incorporate reinforcement learning from human feedback (RLHF) to improve alignment with company policies.
- E
Exclude data governance considerations since they do not impact model fine-tuning.
Show answer and explanation
Correct answers: A, C, D
Explanation
Preparing data for fine-tuning requires ensuring the data is representative of real-world scenarios, properly labeled for the tasks the model must perform, and aligned with organizational goals using techniques like RLHF. Data governance is also critical to ensure ethical and legal compliance while avoiding random, unfiltered data that may introduce errors or biases.
- A. Correct.
Ensuring the dataset includes representative examples is critical to preparing training data that reflects real-world use cases and improves the model’s performance.
- B. Incorrect.
Randomly sampling unstructured data without filtering can introduce noise, irrelevant information, or biases, which can negatively impact the fine-tuned model.
- C. Correct.
Data labeling helps the model learn specific tasks such as identifying intents and sentiments, which are essential for a customer service chatbot.
- D. Correct.
Reinforcement learning from human feedback (RLHF) helps fine-tune the model to align with human expectations and specific company policies, making its responses more accurate and trustworthy.
- E. Incorrect.
Excluding data governance considerations is incorrect as governance ensures the data is compliant, unbiased, and appropriate for use.