How does fine-tuning differ from pretraining?
Why D is the answer
It adapts an existing model with focused data
Pretraining builds general capability, and fine-tuning adapts it.
In detail
- Fine-tuning adapts an existing model with focused data.
- Pretraining builds general capability.
Use case
An illustrative scenario, not a real incident.
A company adapts a general model to its support style through fine-tuning, rather than building one from scratch.
See it in different ways
flowchart LR A["Collect and tokenize"] --> B["Pretraining"] B --> C["Fine-tuning"] C --> D["RLHF"] D --> E["Evaluation"]
stateDiagram-v2 [*] --> Tokenize Tokenize --> Pretraining Pretraining --> FineTuning FineTuning --> RLHF RLHF --> Evaluation Evaluation --> [*]
flowchart LR S0["Pretraining, general"] --> S1["Fine-tuning, focused"] --> S2["Adapted model"]
Key point: Fine-tuning adapts an existing model.
Why the other options are wrong
- A. It builds all of the model's weights from random values
- That is pretraining.
- B. It removes the need to evaluate the resulting model
- You still evaluate.
- C. It only changes the prompt and never changes weights
- It does change weights.