Answers need facts that change every day. What is usually better than fine-tuning?
Why D is the answer
Retrieval from a current source
Fine-tuning fixes knowledge in the weights. Retrieval stays current.
In detail
- Retrieval stays current, and fine-tuning fixes knowledge in the weights.
- Tuning suits stable behavior.
Use case
An illustrative scenario, not a real incident.
Prices change daily. Retrieval reads the current list, while fine-tuning would freeze old facts.
See it in different ways
flowchart TD
A{"Is the need new knowledge?"} -->|Yes| B["RAG"]
A -->|"No, it is behavior or style"| C{"Is the budget limited?"}
C -->|Yes| D["LoRA or other PEFT"]
C -->|No| E["Full fine-tuning"]stateDiagram-v2 [*] --> Need Need --> Retrieval: facts that change Need --> FineTuning: stable style or format FineTuning --> Lora: limited GPU budget FineTuning --> FullTuning: budget allows it Lora --> EarlyStopping: small dataset
flowchart TD
Q{"What must the model know?"}
Q -->|"Daily facts"| A0["Retrieval"]
Q -->|"A stable style"| A1["Fine-tuning"]Key point: Retrieval stays current.
Why the other options are wrong
- A. A daily full retraining of the model
- Daily retraining is far too costly.
- B. A higher temperature, for fresher output
- Temperature adds no facts.
- C. Quantizing the model to eight bits
- Quantization adds no facts.