Process Reward Step Values

Problem

Compute the discounted process-reward value at every step.

Walking backward, V[t] = step_rewards[t] + gamma * V[t + 1] with the value after the last step equal to 0. Return the list of values; an empty episode returns [].

Hint

A process reward model scores each reasoning step, not just the final answer.

Test cases

case 1

in: [[1,1,1],0.5]

expected: [1.75,1.5,1]

case 2

in: [[1,1,1],0]

expected: [1,1,1]

case 3

in: [[0.5],0.9]

expected: [0.5]

case 4

in: [[],0.9]

expected: []

Starter codePython

def prm_step_values(step_rewards, gamma):
    # Your code here
    pass

4 test cases · run your code against them in the editor