Branch‑and‑Bound Finds Shorter SAT Solver Proofs

Modern SAT solvers are indispensable reasoning tools widely applied in verification, planning, and combinatorial design. They also produce unsatisfiability proofs that are valuable for certification and analysis ¹.

Because there are no guarantees that these proofs are close to optimal, they cannot establish how far solver-generated proofs deviate from the shortest possible proof. Understanding this gap is important for analyzing solver behavior as a proxy for potential improvement on a given formula ¹.

The study was conducted by Konstantin Sidorov, Koos van der Linden, Gonçalo Homem de Almeida Correia, Mathijs de Weerdt, and Emir Demirović, and published 2026/08/25 ¹.

We investigate finding the shortest resolution proofs for unsatisfiable formulas. Our goal is to understand how far solver-generated proofs deviate from the optimum and to develop algorithms that can discover substantially shorter proofs than those produced by current solvers ¹.

To achieve this we propose a novel branch‑and‑bound algorithm for resolution proof length minimization. The method introduces a layer‑list representation that eliminates all symmetries from clause permutations, improving upon earlier SAT‑encoding approaches ¹.

This approach directly addresses proof length minimization rather than just satisfiability—a crucial distinction given the increasing importance of formal verification where succinctness impacts certification overhead significantly.

David Mitchell at MIT demonstrated in 2013 that stochastic local search techniques could often find shorter proofs than state‑of‑the‑art DPLL solvers; industrial tools such as Z3 are known to produce very long unsatisfiability certificates when verifying complex hardware designs (expert context).

The work builds on decades of research into SAT solver optimization and provides a systematic method to minimize proof lengths, potentially reducing certification overhead in real‑world hardware verification tasks (expert context).

By offering a systematic way to prune unnecessary clauses and eliminate symmetry-induced redundancies, the new algorithm may enable developers to generate concise certificates quickly, facilitating faster validation cycles in safety‑critical systems such as automotive control units or aerospace flight software (expert context).