Knowledge Injection
Domain: Prompt Engineering
🎯 Motivation
Providing specific facts or domain knowledge in the prompt that the model doesn’t have from training, ensuring accurate domain-specific answers.
📋 Concrete Example
‘Context: Our school’s grading scale is A(90+), B(80-89), C(70-79), D(60-69), F(<60). Based on this scale, grade these scores: [scores].’ The injected knowledge ensures correct grading.
🔗 Analogy
Giving a substitute teacher the lesson plan - they don’t know the class’s progress (missing knowledge), so you inject it via the lesson plan. Knowledge injection gives the LLM the specific facts it needs.