Logit Bias Application

Problem

Apply per-index logit biases and return the new logits.

For each k add biases[k] to logits[indices[k]] and return a new list; the original must not be modified. Empty index lists return a copy of the logits.

Hint

Bias vectors let you steer generation without retraining the model.

Test cases

case 1

in: [[1,2],[1],[5]]

expected: [1,7]

case 2

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

expected: [-1,1]

case 3

in: [[],[],[]]

expected: []

Starter codePython

def apply_logit_bias(logits, indices, biases):
    # Your code here
    pass

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