Algorithm · Amazon · Medium
Requirements You are asked to finish a coding notebook that centers on the fundamentals of GRPO. The tasks are: Complete the coding notebook focused on GRPO fundamentals. Use NumPy operations inside the notebook. Cover how GRPO assigns process rewards. Cover how GRPO assigns shape reward. Interview Notes This was the third interview in a three-round onsite loop for an AGI Applied Scientist position. The exact meaning of shape reward, the reward formulas, the tensor shapes,…