Language-Conditioned Reward

Problem

Compute a language-conditioned reward as the cosine similarity between the language embedding and the state embedding. A zero-norm vector gives 0.0.

Hint

Cosine similarity makes the reward scale-free with respect to embedding magnitude.

case 1

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

expected: 1

case 2

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

expected: 0

case 3

in: [[1,1],[2,2]]

expected: 0.9999999999999998

case 4

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

expected: 0

Starter codePython

import math
def language_conditioned_reward(language_embedding, state_embedding):
    # Your code here
    pass

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