If you got problems with WA 20 (new test), just know - it is because of precision issue. Not use double or long double - only long integer arithmetic   or   integer modular arithmetic.

P.S.
   authors of new tests - it is not good to add tests to the OLD task after 23 years of time passed. It is so strange. Why authors of solutions must return to the OLD task and resolve it again???????

Thanks LLI! By the way, I used integer modular arithmetic in the first place, and still got WA 20 after rejudge.

After some debugging I realize I checked the linear independency incorrectly (in particular, I did the row echelon form wrongly). Here is the test case helped me find out the bug:

Input:
5 4
1 2 3 4
2 4 5 9
3 6 7 1
1 1 1 9
3 2 1 9
1
2
3
4
5

Expected output:
10
1
2
3
4