We introduced Sudoku as a CSP to be solved by search over

partial assignments because that is the way people generally undertake solving Sudoku problems. It is also possible, of course, to attack these problems with local search over complete silence how well would a local Solver using the min-conflicts heuristic do on sudoku problems?

It can be a really good approach to use a local solver using the min conflicts heuristic in solving sudoku problems. It will work better actually. In this process, the value chosen is the value with the minimum conflicts. This is the general way a normal person would also tackle this problem. By this approach, if we keep taking the values with minimum conflicts the sudoku puzzle can be solved with a better performance.