Effectiveness of Sparse but Informative Human Feedback in RLHF

A team is training a language model to write compelling short stories. They decide to provide a single quality score (reward) only after an entire story is completed, rather than scoring each sentence as it is generated. Which of the following statements best analyzes the primary justification for this training strategy?

AI Training Feedback Strategy

When training a language model to write a multi-paragraph summary of a document, providing a reward after each correctly structured sentence is generated is a more practical and effective approach than providing a single, holistic reward based on the quality of the final, complete summary.