how to ruin another model: reflection 70b

reflection 70b is less based than llama 3.1 70b.

while trying to "fix" reasoning capabilities they rekt the truthfulness.