Free 🧰 Practical LLM Lessons

LLM Lab 7: Context and Grounding

Give the model the right evidence and practice grounding answers in provided context.

Published January 8, 2026

Previous

LLM Lab 6: Chain-of-Thought Builder

Next

LLM Lab 8: Evaluate the Output

Grounding beats guessing

If the model lacks the facts, it will improvise. Grounding means supplying the evidence in the prompt.

Can You Guess?

You paste meeting notes and ask: 'Who owns the Q1 launch? Cite the source.' The model invents 'Alex'. What should you change first?

Build a grounded prompt

Follow the steps and watch each checkpoint light up as you progress.

1

Step 1: Provide the source

Paste or link to the text the model should trust.

2

Step 2: Limit scope

Tell it to answer only from that source and to say when missing data.

3

Step 3: Ask for citations

Request quotes or line numbers to force grounding.

4

Step 4: Check for drift

If it hallucinates, remind it to refuse when evidence is absent.

Try this now

Force evidence

  • Paste a short policy document.
  • Ask a question and require a quoted answer plus the exact sentence.
  • If the model guesses, tighten the instruction: “If not found, say ‘not in the doc’.”

    • Key Takeaways

    • ✓ Grounded prompts supply the evidence.
    • ✓ Require citations to reduce hallucination.
    • ✓ When data is missing, instruct the model to say so.

    Quick poll

    What brought you to this free lesson?

    Lesson: LLM Lab 7: Context and Grounding

    ← Previous Lesson LLM Lab 6: Chain-of-Thought Builder Next Lesson → LLM Lab 8: Evaluate the Output
    Back to Library Unlock More Lessons