HardNLPnlp-157

Best Span Selection

Problem

Select the best QA span with a maximum length constraint: maximize start_logits[i] + end_logits[j] over j >= i and j - i + 1 <= max_len.

Return [i, j, score] with the lowest i and then lowest j winning ties, or an empty list when no span is valid.

Hint

Only replace the best on a strictly greater score so ties keep the earliest span.

case 1

in: [[0.5,0.1],[0.4,0.2],2]

expected: [0,0,0.9]

case 2

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

expected: [0,0,0]

case 3

in: [[1,0,0],[0,0,2],1]

expected: [2,2,2]

case 4

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

expected: []

case 5

in: [[],[],2]

expected: []

Starter codePython

def best_span_selection(start_logits, end_logits, max_len):
    # Your code here
    pass

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