EZSMTV3 Brings SMT Muscle to Answer Set Programming's Hardest Constraints

A research team led by Yuliya Lierler has released EZSMTV3, a solver that pushes Constraint Answer Set Programming (CASP) further into territory once reserved for dedicated theorem provers. The system is described in a paper submitted to arXiv on July 15, 2026, and currently under review for publication in Theory and Practice of Logic Programming (TPLP).

CASP frameworks exist to solve problems that plain Answer Set Programming (ASP) struggles with: tasks involving numerical constraints, scheduling limits, or continuous variables that don't map cleanly onto discrete logic rules. EZSMTV3 tackles this by fusing ASP with Constraint Processing and Satisfiability Modulo Theories (SMT), letting a single framework reason over logical rules and arithmetic constraints simultaneously.

The system builds directly on EZSMT+, an earlier solver from the same research lineage, but expands what users can express in their input programs. According to the paper, EZSMTV3 supports a noticeably richer input language than its predecessor, giving developers more flexibility in how they encode real-world constraints without falling back on workarounds or manual reformulations.

Mixed-Domain Reasoning and Optimization

One of the more consequential upgrades is EZSMTV3's ability to handle mixed-domain constraints, meaning problems that combine integer and real-valued variables within the same program. This matters for applications like resource allocation, planning, or configuration problems where some quantities are naturally discrete (counts, indices) and others are continuous (costs, measurements, time). Forcing everything into one domain often means sacrificing precision or expressiveness; EZSMTV3 avoids that trade-off.

The framework also adds support for optimization through weak constraints, a mechanism in ASP that lets solvers search not just for any valid answer but for the best one according to a defined preference or cost function. Combined with mixed-domain support, this positions EZSMTV3 as a tool suited for constraint satisfaction problems that also require ranking or minimizing among multiple valid solutions.

Built on Established SMT Solvers

Rather than implementing its own theorem-proving engine from scratch, EZSMTV3 delegates the heavy lifting of satisfiability checking to established SMT solvers: CVC5, YICES, and Z3. This design choice lets the CASP layer focus on translating ASP-style problems into a form these solvers can process, while benefiting from years of optimization work already baked into those backends.

To validate the approach, the researchers benchmarked EZSMTV3 against other CASP systems, including CLINGCON, CLINGO[DL], and CLINGO[LP]. These represent some of the most established tools in the field, each taking a different approach to bridging ASP with numeric or constraint-based reasoning. The comparison positions EZSMTV3 within an active and competitive corner of logic programming research, one where expressiveness, solver performance, and ease of encoding are all contested ground.

For practitioners working on planning, verification, or configuration problems that resist purely discrete formulations, EZSMTV3's combination of expanded input syntax, mixed-domain support, and weak-constraint optimization offers a more flexible alternative to prior CASP tooling, backed by mature SMT infrastructure rather than bespoke solving logic.