Which statement about model cards is true?

Which statement about model cards is true?

• A. They are primarily used to track deployment uptime.
• B. They describe model purpose, limitations, and performance.
• C. They are optional and rarely used.
• D. They focus only on code structure.

Answer: (B)

Model cards provide a clear, structured description of a model's intended use, its limitations, and its performance. They are written to communicate to stakeholders what the model is designed to do, who it’s for, and under what conditions its results are reliable. A model card typically includes the purpose and scope of the model, the data it was evaluated on, performance metrics across relevant scenarios or subgroups, known limitations and potential failure modes, and guidance on safe or appropriate use. This kind of documentation supports transparent decision-making, risk assessment, and responsible deployment, helping teams anticipate where the model may misbehave and what safeguards to apply. So the statement that model cards describe model purpose, limitations, and performance is true. They are not about tracking deployment uptime, they’re not merely optional or rarely used, and they aren’t solely focused on code structure.

Model cards provide a clear, structured description of a model's intended use, its limitations, and its performance. They are written to communicate to stakeholders what the model is designed to do, who it’s for, and under what conditions its results are reliable. A model card typically includes the purpose and scope of the model, the data it was evaluated on, performance metrics across relevant scenarios or subgroups, known limitations and potential failure modes, and guidance on safe or appropriate use. This kind of documentation supports transparent decision-making, risk assessment, and responsible deployment, helping teams anticipate where the model may misbehave and what safeguards to apply.

So the statement that model cards describe model purpose, limitations, and performance is true. They are not about tracking deployment uptime, they’re not merely optional or rarely used, and they aren’t solely focused on code structure.