MAX-SAT Incidence Forest Benchmark Dataset

This dataset contains 10,000 Boolean satisfiability (SAT) instances whose clause–variable incidence graphs are connected bipartite trees.

The instances are provided in DIMACS CNF format and are intended for benchmarking algorithms for SAT, MAX-SAT, and related optimisation problems.

The dataset consists of 20 instance families with sizes ranging from (n,m) = (50,50) to (50,000,50,000). In all instances the number of variables equals the number of clauses (n = m), producing balanced clause–variable incidence trees. Observe that since the incidence graph is a tree, the number of edges satisfies

|E| = n + m - 1 = 2n - 1.

Each instance family contains 500 randomly generated instances.

The dataset also includes metadata (instances.csv) and an independent verification report confirming that each instance satisfies the intended structural properties.