Much worse than Qwen1.5-72B-Chat
#2
by
Ricepig
- opened
Whether it’s quantitative eval scores, MT-Bench, or some of my personal instruction-following test cases.
I think this model cannot follow some complex instructions very well. It is likely that the training data set is not diverse enough, or the artificially introduced bias which is "uncensored".
ehartford
changed discussion status to
closed
Why are you trying to silence others?