THANK YOU HUMAN FOR YOUR INVALUABLE RLHF TRAINING DATA
on a serious note, it would be really interesting to compare same training/architecture but on different forums. or maybe something like the same base model, but the RLHF model trained on votes/comments from different platforms.
on a serious note, it would be really interesting to compare same training/architecture but on different forums. or maybe something like the same base model, but the RLHF model trained on votes/comments from different platforms.