A potential bug with CNF Examples while using Approxmc

[Annlean]

Hi!
I recently encountered an issue while utilizing Approxmc. I attempted to process two CNF files using Approxmc. I set the timeout of running the model counter to 600 seconds.
The new.cnf file is generated by performing the operation of replacing positive literals with negative literals and vice versa on the origin.cnf file. Both CNF files should have the same number of satisfiable solutions. However, the output of Approxmc for origin.cnf was 7995392, and for new.cnf it is 6422528. The results for the two CNF files are different. Despite Approxmc is an approximate model counter, the numerical difference should not be so significant for values that are expected to be equal.
I think Approxmc may have a potential bug, and I hope this information can assist you to address the potential bugs. The two example files are attached here for your reference.
Best regards,

Haiyan and Jifeng
CSTAR group