All the instances provided were generated by our instance generators
presented in the workshop.  All the instances are for 3SAT and follows
the DIMACS format.

Merits of our generator is that it can generate yes-instances and
no-ones independently for wide ranges.  So our primary role is
probably to provide such instances that the conventional random
generation can hardly generate.  At the same time, our generator runs
in randomized fashion, so that it essentially different from those
generated in some deterministic fashion or those translated from other
problems.

As a result, the following set of instances include (i) no-instances
at low clause/variable ratios and (i) yes-instances that however has
only one solution (assignment making the predicate true).  (ii) was
generated by a newly developed generator whose basic idea is the same
as our no-instance generator and would be useful to test some type of
incomplete algorithms.

The instances are named as aim-xxx-y_y-zzzz-j where
(i) all the names has "aim" at the beginning (stands for Asahiro, Iwama
    and Miyano).
(ii) next xxx shows the number of variables, one of 50, 100 and 200,
(iii) y_y shows the clause/variable ratio y.y, including 1.6, 2.0 for
      no-instances and 1.6, 2.0, 3.4 and 6.0 for single-solution yes-
      instances,
(iv) zzzz is "no" or "yes1", the former denoting a no-instance and the
     latter a single-solution yes-instance,
(v) the last j means simply the j-th instance at that parameter.

For each parameter, we include four instances, e.g., aim-100-2_0-no-1
through aim-100-2_0-no-4, which are four no-instances with 100 variables
and 2.0 clause/variable ratio.  According to CSAT by Dubois, et al,
the first and third instances are harder than the second and forth
ones.

Kazuo Iwama, Eiji Miyano and Yuichi Asahiro
Dept of Computer Science and Communication Engineering
Kyushu University
Fukuoka 812, Japan
+81-92-641-1101 (ex. 5407) +81-92-632-5204 (fax)