6 ms·
Tuning for qualitative outcomes is pretty much solved via RLHF/DPO (what this post calls "preference tuning"). Right?
by researchers 2y ago
Tuning for qualitative outcomes is pretty much solved via RLHF/DPO (what this post calls "preference tuning"). Right?