C.W.K.
Stream
Quiz · 6 questions

🧠 Reasoning-Oriented Models

Same backbone, different training and inference — thinking longer for harder problems

Level 0Scout
0 XP0/41 lessons0/12 achievements
0/100 XP to next level100 XP to go0% complete

Quiz

01What is the MOST accurate description of modern reasoning models?
Hint
Same backbone, different training and inference.
02What is 'test-time compute scaling' (TTS)?
Hint
Compute spent at *inference* time, not training time.
03DeepSeek-R1-Zero demonstrated something remarkable. What was it?
Hint
Pure RL, no SFT — and the model still figured out how to reason.
04Why is reasoning-model 'overthinking' a practical problem?
Hint
More thinking helps when there's something to think about.
05Which RL algorithm did DeepSeek introduce that eliminates the need for a separate critic network in reasoning RL?
Hint
G stands for 'Group'; the trick is comparing samples to *each other*, not to a learned baseline.
06The same Claude or Qwen3 checkpoint can serve both 'fast' and 'thinking' modes via a configuration flag. What does this prove?
Hint
If the weights file is the same, only inference can be different.
Spotted a bug or have feedback on this page?Report an Issue

Comments 0

🔔 Reply notifications (sign in)
Sign inPlease sign in to comment.

No comments yet — be the first.