We develop a tabu search algorithm for the quadratic multiple knapsack problem.
The algorithm incorporates a hybridization scheme combining both feasible and infeasible local searches.
Out of the 60 instances we obtain 10 strictly new best solutions.
Non-parametric tests emphasize the effectiveness of the proposed approach.