SMT-COMP 2025 Benchmarks

This submission contains all benchmarks that were used in SMT-COMP 2025. The benchmarks are divided to five archives, according to the corresponding track

• single query,
• incremental,
• unsat core,
• model validation,
• parallel.

The benchmarks are a selection of benchmarks from SMT-LIB repository that were scrambled by the official SMT-COMP scrambler with the seed 757067271.

Each archive contains subdirectories for each used logic, i.e., single_query/ABV. For each benchmark, the subdirectory contains two files:

• ID_NAME.yml, where ID is the id of the benchmark used by our internal scripts and NAME is the mangled name of the original benchmark (including its family). For example:  single_query/ABV/21_ABV_20190429-UltimateAutomizerSvcomp2019_cs_szymanski_true-unreach-call.i_6.yml.The file contains additional metadata for the execution using BenchExec such as the expected status, full name of the original benchmark, and, most importantly, the corresponding smt2 file. This file is specified by the key input_files, e.g., input_files: 'scrambled380969.smt2'.
• scrambledNUMBER.smt2, where NUMBER is a random number assigned during the scrambling process, e.g., single_query/ABV/scrambled380969.smt2. This is the file that is given as an input to participating SMT solvers. The connection between the scrambledNUMBER.smt2 file and the original benchmark is specified only in the YML file described above, which is used only by BenchExec and not available to the participating SMT solvers. This ensures that the SMT solvers do not have access to the names of the original benchmarks.