undefined | Better HN

0 pointsJackson__2y ago0 comments

You might be thinking of unofficial LLaMA finetunes such as Alpaca, Vicuna, etc.

LLaMA 1 was a base model without any safety features in the model itself.

0 comments

3 comments · 1 top-level

Base LLaMAv1 would refuse to answer certain questions. It wasn't as aggressive as OpenAI models or the safety aligned finetunes, but some kind of alignment was there.

astrange2y ago

Normal training content has "alignment". It's not going to instantly be super racist and endorse cannibalism if it's "unaligned".

brucethemoose22y ago

It very specifically mentioned something about LLaMA not being trained to answer that in the response.

Again, its extremely minimal, but I think it picked something up from the Llama info facebook inserted.

1 more reply

j / k navigate · click thread line to collapse