AWS AI Practitioner
A data scientist wants to improve the domain knowledge of a pre-trained LLM without using labeled data. Which training technique should the data scientist use?
A
Fine-tuning
B
Continued pre-training
✓ Correcta
C
Reinforcement learning from human feedback (RLHF)
D
Transfer learning
Explicación
Continued pre-training uses unlabeled domain-specific data to extend a pre-trained LLM's knowledge in a particular domain. Unlike fine-tuning which requires labeled data, continued pre-training can improve domain knowledge using raw text data without labels.