Looks like default reading in python3 is incorrect. I've tried the following code with test case #6

#!/bin/python3

import sys

if __name__ == "__main__":
    n, m, x, k = input().strip().split(' ')
    n, m, x, k = [int(n), int(m), int(x), int(k)]
    a = list(map(int, input().strip().split(' ')))
    b = list(map(int, input().strip().split(' ')))
    # Write Your Code Here
    print(n, m, x, k)
    print(len(a),len(b))

the output on my computer is:

1000 2000 1000 206630
588 595

instead of expected

1000 2000 1000 206630
1000 2000

Disappointed that the C++ boilerplate solution still uses int vars, when the constraints obviously required something else.

Awarding 20% points because I've discovered the bug in the boilerplate?? I expected better.

With respect: your problems are interesting, and I suspect good examples of exploring problem spaces. But I believe your descriptions are too terse. When I have fifteen minutes to solves a problem (HackHour 24), but still cannot understand the simpler problems after 5 minutes? Agree, I probably need to study the algorithm space more ... and am happy to research. But in 10 minutes? It becomes not a test of skill, but of specific domain knowledge. I think the questions are obfuscated, and ask that you vet them a little more.

generate_list function in Python would be:

def generate_list(A, B, n, m, x):
    L = []

    for i in xrange(min(n, m - x)):
        for j in xrange((i + x), m):
            L.append(A[i]*B[j])
            
    return L

Do I have to optimize this part as well? Or leave it as it is? I only passed 11 test cases using sort and heapq (I changed append to heappush).

I still have a question: is there an error in "explanation 0"? why is list B iterated from 3rd element instead of 2nd? )

Using min-heap and then popping doesn't make the cut either. ;-(