Stripped binaries in dataset cause bad ground truth comparisons

[aeflores]

I have noticed that several binaries in the dataset are stripped:
e.g.

mthumb_executables/clients/gcc_O2/scp
mthumb_executables/clients/gcc_O2/sftp
mthumb_executables/clients/gcc_O2/ssh
mthumb_executables/clients/gcc_O2/ssh-add
...

since the comparison scripts use symbols to exclude functions added by linker (https://github.com/junxzm1990/x86-sok/blob/master/compare/compareInstsArmMips.py#L504), this can cause problem in these binaries.

For example, in mthumb_executables/clients/gcc_O2/scp, the address 0x28c4 does not seem to belong to the ground truth, however, that is the entrypoint address (the entrypoint is 0x28c5 according to readelf since this is a thumb binary).
In other words, the _start function is missing from the ground truth and it cannot be excluded because the binary is stripped.
Do you have any advice on how to deal with these cases?

[bin2415]

We have recompiled the binaries related to ssh in Thumb mode.

The binaries are shared in google drive