What is fine-tuning in the context of large language models?
Fujitsu aptitude question, verified with a worked answer. Free to practise - no sign-up.
What is fine-tuning in the context of large language models? What is fine-tuning in the context of large language models? © Adjusting the model's physical hardware for better performance - A) The process of training the model on specific tasks or datasets - B) Modifying the model's core architecture - C) Converting the model's output into different languages
Show answer & explanation
Fine-tuning is a supervised learning process where a pre-trained model is further trained on a smaller, specialized dataset to adapt its general knowledge to a specific domain or task.
Step-by-step Derivation:
Step 1: Analyze the definition of Large Language Model (LLM) training. LLMs typically undergo two main phases: Pre-training (on a massive, general corpus to learn language patterns) and Fine-tuning (on a curated dataset for specific goals).
Step 2: Evaluate Option B: Modifying core architecture refers to changing the number of layers, attention heads, or the model type (e.g., from Transformer to something else), which is not 'fine-tuning' but 'architectural redesign'.
Step 3: Evaluate Option C: Converting output into different languages is 'Translation' or 'Inference-time processing', not a training process.
Step 4: Evaluate Option A: Training a pre-trained model on a task-specific dataset (like medical records or legal documents) to refine its weights for that specific use case matches the technical definition of fine-tuning.