Quiz · 6 questions
🚫 What Is NOT an Architecture
RAG, tools, agents, long context, speculative decoding — all wear architecture's coat, none of them are
Level 0Scout
0 XP0/41 lessons0/12 achievements
0/100 XP to next level100 XP to go0% complete
Quiz
01A company claims their AI has 'built-in retrieval capabilities'. What does this most likely mean?
Hint
If it can be removed by skipping a few lines of orchestration code, it lives outside the model.
02Speculative decoding produces output that is:
Hint
If two methods produce the *same* outputs, only the time differs.
03Which of the following IS a genuine backbone architecture difference (axis 1)?
Hint
Backbone = what is wired. Wiring lives in config.json.
04An 'AI agent' framework like LangChain wraps an LLM. What changes inside the model when you run it as an agent?
Hint
If the framework can be swapped out without retraining, it lives outside the model.
05GQA, MQA, and MLA are all variants that change which property of attention?
Hint
What gets shared, not whether attention happens.
06A model claims to have a '1M context window'. What is the SAFEST thing to assume?
Hint
Capacity ≠ capability.
Comments 0
🔔 Reply notifications (sign in)Sign in — Please sign in to comment.
No comments yet — be the first.