An AI team is building a checklist for selecting a foundation model. As specified in the question, performance and cost have already been evaluated. Which factors should be considered beyond performance and cost? (Choose TWO.)
Choosing TWO factors to consider beyond performance and cost in model selection.
Whether security and compliance requirements can be met
Correct. Whether the sensitivity of the data handled and regulations (storage location, encryption, auditing, and so on) can be met is an important selection factor beyond performance and cost.
The supported modality
Correct. The modality, such as whether the model handles only text or also images and audio, is an important factor in choosing a model that fits the use case.
Inference latency
Latency is an important selection criterion, but it is a performance metric.
This question asks for factors 'beyond performance and cost,' so latency, being on the performance side, is out of scope and incorrect.
Price per input and output token
Price per token is an important selection criterion, but it is a cost metric.
This question asks for factors 'beyond performance and cost,' so it is out of scope and incorrect.
The number of epochs set during training
The number of epochs is an internal setting the provider used when training the model, not a factor users consider in model selection.
It does not serve as a selection criterion, so it is incorrect.
In selecting a foundation model, in addition to performance (accuracy, latency) and cost, 'whether security and compliance requirements can be met' and 'the supported modality (text, images, and so on)' are important considerations (others include context window, customizability, and licensing).