3 ms·
> which is very likely to be bad for reasoning capabilities. As opposed to our aligned models which refuse to answer questions with controversial answers?
by silver_silver 2y ago
> which is very likely to be bad for reasoning capabilities.
As opposed to our aligned models which refuse to answer questions with controversial answers?