Reinforcement Learning from Human Feedback

3 selectedDifficulty 5-73 unseenView topic

Saved practice

Keep this quiz in your learner record

Answers count toward your profile, review queue, and next-topic suggestions. You can also use the quick practice below.

IntermediateNew

0 answered

2 intermediate1 advancedAdapts to your performance

Question 1 of 3

120sintermediate (5/10)compare

RLHF typically trains a reward model on pairwise human preferences before policy optimization. Why preferences rather than absolute ratings?