Follow this link to a summary table of results and links to problem data.  The set partitioning problem is a minimization, however, our tabu search codes are maximizations, therefore the set partitioning Q matrix is modified accordingly.  To guarantee feasibility, a new variable is appended to each row, but is given a 2*max_coefficient penalty in the objective function. 
The 3 formats provided are cplex .lp, upper triangular Q matrix in row/col/value format, and the full Q matrix.  An example of a small problem with Penalty = 50 is given below.
 

1)  the cplex .lp format e.g.

\\ Set partitioning with 5 vars
Minimize
 + 1 x_1 + 3 x_2 + 7 x_3 + 20 x_4 + 20 x_5
 st
\\ Sum x_i = 1  for j = 1 to m = 2 constraints
 + x_1 + x_2 + x_4 = 1
 + x_1 + x_3 + x_5 = 1

Binaries
x_1   x_2   x_3   x_4   x_5   
end

2) the row column value format describing the upper triangular portion of the Q matrix (with the first two rows describing Palubeckis multi-start tabu search parameters) with the diagonal elements doubled (e.g. 2*(50-1) = 98)

1 5 100 4000 4000
5 17

1 1 98
1 2 -50
1 3 -50
1 4 -50
1 5 -50
2 1 -50
2 2 94
2 4 -50
3 1 -50
3 3 86
3 5 -50
4 1 -50
4 2 -50
4 4 60
5 1 -50
5 3 -50
5 5 60

and 3) the full Q matrix where the first row and the last three rows are Kochenberger tabu search parameters. 

xqx_3_2_30_0.dat
  5 0
99 -50 -50 -50 -50
-50 47 0 -50 0
-50 0 43 0 -50
-50 -50 0 30 0
-50 0 -50 0 30
   3 5 5 1 0 0 0
   7 3 5 500 -99 100 100 1.0 0.01 0.0
   99 1371 1 1