Using scribmol, I drew but-2-ene (cis-conformation) in chemdoodle. The corresponding 3D molecule is created by Jmol.
"show smiles" gives : C/C=CC
when transfered to cactus for optimization, the resulting 3D file is in trans configuration (the 2D mol file shows a crossed double bond)
Sometimes (I can't find a reproductible scenario), when I draw the same molecule directly in Jmol, the "show smiles" gives : C(/C)=C/C which leads to a correct optimization.
In trans configuration, the results are quite similar : "show smiles" gives C(/C)=CC which leads to a crossed double bond with cactus, but provides a molecule in trans configuration (I suppose it is the default)