exploration
10 minTesting the Semantic Baseline
semantics
prediction
n-grams
Tools:Google AI Studio
This activity explores how Large Language Models (LLMs) predict text based on meaning (semantics) rather than just statistical frequency (n-grams). You will test the model's ability to understand 'vibes' versus precise details.
Step 1 of 4
Getting started
Navigate to ai.dev and start a New Chat.
Generate Text
Ask Gemini to write a short story (e.g., 'Write a paragraph about a student named Elara arriving late to class') or paste in a text you know well.
Test Understanding
Ask questions about the text to see how it understands it. Try asking it to fill in blanks (cloze test) or identify specific details (e.g., 'Why was the teacher sarcastic?').
Observe
Note where the model succeeds in capturing the meaning and where it might fail on precise details.