179 questions
No questions match those filters.
← All questions
Training & Fine-tuningMedium· RLHF· Reward Modeling· Reinforcement Learning
What is the difference between outcome and process rewa...
This is one of the questions in the full AI/ML interview bank. Pro unlocks all 1789 questions; Premium includes the same bank plus the highest daily Practice limit.
See plansOutcome rewards are deterministic and costless, requiring no learned scorer. In contrast, process rewards provide a denser gradient signal but necessitate expensive step-level annotation or a learned PRM that may introduce proxy noise. PRMs should be reserved for scenarios where difficulty-filtered outcome rewards yield near-zero group pass rates across the training set.