Question 31
Domain 3: Applications of Foundation ModelsA company wants to keep its foundation model (FM) relevant by using the most recent data. The company wants to implement a model training strategy that includes regular updates to the FM. Which solution meets these requirements?
Correct answer: B
Explanation
Continued Pre-training keeps a foundation model current by continuing training on new data, so it can absorb “the most recent data” without starting from scratch. This matches a regular-update strategy because the model is repeatedly updated as new information arrives, unlike one-time pre-training or fine-tuning on a fixed dataset.
Why each option is right or wrong
A. Batch learning
Batch learning trains on a fixed dataset in discrete runs, not ongoing FM updates.
B. Continued Pre-training
Continued Pre-training is the training strategy that keeps a foundation model current by continuing to train it on newly arriving data, which fits the requirement for regular updates from the latest weekly or streaming corpus. Unlike one-time pre-training, it does not restart from scratch; instead, it incrementally incorporates new information so the model remains relevant as the data distribution changes over time.
C. Static training
Static training is a one-time process; it does not incorporate recurring new data.
D. Latent training
Latent training is not the standard term for updating a foundation model with fresh data.