SemML 2.0 Outperforms State-of-the-Art LTL Synthesis Tools
The latest iteration of the tool, SemML 2.0, facilitates the synthesis of reactive systems from linear temporal logic (LTL) specifications, addressing a longstanding challenge in the design of safety-critical systems. In the SYNTCOMP synthesis competition dataset, it surpasses all leading tools, such as Strix, LtlSynt, and its predecessor, SemML, by solving a greater number of instances at a significantly quicker pace. While SemML 2.0 adheres to the traditional automata-theoretic method, it also incorporates partial exploration and machine-learning techniques for enhanced efficiency, alongside various heuristics and refinements of established algorithms to derive compact solution representations. The tool can model systems as either Mealy machines or AIGER circuits. The research paper is accessible on arXiv with the identifier 2604.24102.
Key facts
- SemML 2.0 is a tool for synthesizing reactive systems from LTL specifications.
- It outperforms Strix, LtlSynt, and the previous SemML version on SYNTCOMP data.
- It solves more instances faster than other tools.
- It uses automata-theoretic approach, partial exploration, and machine-learning guidance.
- Systems are represented as Mealy machines or AIGER circuits.
- The paper is on arXiv: 2604.24102.
Entities
Institutions
- arXiv