Databricks Generative AI Engineer Associate Question 166
Select 3You are tasked with selecting a generative AI model from a model hub to perform text summarization on large legal documents. The model should minimize bias and support long input sequences. Which model card metadata would be most relevant to review for this task?
- A
Input context length supported by the model
- B
Training dataset description and sources
- C
Model's fine-tuning capabilities
- D
Number of model parameters
- E
Evaluation metrics for summarization tasks
Show answer and explanation
Correct answers: A, B, E
Explanation
To select an optimal model for text summarization on large legal documents, it is important to consider the input context length (to handle long inputs), the training dataset (to ensure diversity and minimize bias), and the evaluation metrics specific to summarization tasks. These factors directly influence the model's suitability for the task at hand.
- A. Correct.
Input context length is crucial for summarizing large legal documents as it determines how much text the model can process in a single input.
- B. Correct.
The training dataset description and sources help assess whether the model has been trained on diverse and unbiased data, which is important for minimizing bias.
- C. Incorrect.
While fine-tuning capabilities are useful, they are not directly relevant to selecting the most appropriate model for summarization based on the metadata.
- D. Incorrect.
The number of model parameters is a secondary consideration as it does not directly indicate the model's performance for text summarization tasks.
- E. Correct.
Evaluation metrics for summarization (e.g., ROUGE scores) provide insight into the model's performance for the specific task of text summarization.