Training Data #5
-
|
Hey, I've read some high-level information on the website and in this repo. However, I am missing the probably most important practical point: does Roost provide various (SFT) data to make own / open source LLMs more safe? Many thanks in advance! |
Beta Was this translation helpful? Give feedback.
Replies: 2 comments
-
|
Hi @stefan-it - we don't provide training data to fine-tune LLMs (or to eval existing models). The reason is: labeled data is tightly coupled to policies and we platforms have widely varying policies. Thanks, |
Beta Was this translation helpful? Give feedback.
-
|
+1 to @vinaysrao1 but @yallaziad did add some sample policies here: #4 |
Beta Was this translation helpful? Give feedback.
Hi @stefan-it - we don't provide training data to fine-tune LLMs (or to eval existing models).
The reason is: labeled data is tightly coupled to policies and we platforms have widely varying policies.
Thanks,