At a high level, what does self-attention do — and when would you fine-tune rather than prompt a foundation model?
Answer it out loud before you open anything. The value of the flags below is in comparing them to what you actually said — including whether you asked about the data before naming a model.
A customer-support platform classifies incoming tickets into 40 routing categories. A prompted hosted LLM gets acceptable accuracy at a cost per ticket that is too high for the volume, and its outputs occasionally drift when the provider updates the model. The team has 300,000 labelled historical tickets.
What it is really testing
Whether the candidate can explain attention without equations but with the mechanism — every token computes a weighted combination of every other token, with the weights learned from content — and can then draw the boundary between prompting, fine-tuning and training a small model from the volume, the cost, the latency, the stability and the labels available.