RLHF in LLMs
The cluster focuses on discussions attributing specific behaviors, personalities, and response styles in large language models to Reinforcement Learning from Human Feedback (RLHF), distinguishing it from base model pre-training.
Activity Over Time
Top Contributors
Keywords
Sample Comments
Are you taking the RLHF into account when you say so?
RLHF is probably the reason for this.
It's almost certainly the RLHF, not the base model.
Read about RLHF, i think you are misunderstanding what this will be used for.
I think it's part of the RLHF tuning as well
He’s talking about RLHF - reinforcement learning with human feedback (the process that trained ChatGPT), the training data for which is not publicly available.And the point is that you don’t need to RLHF as long as you have access to another model that has been trained with RLHF that you can blackbox.
Hasn't RLHF and with LLM feedback been around for years now
RLHF your AI until you like the output?
That's probably more because of RLHF though, they've optimised for certain kind of responses rather than simple model loss on internet text.
It is RLHF if I understand correctly.