4/ Fine-Tuning LLMs for Factuality – fine-tunes language model for factuality without requiring human labeling; it learns from automatically generated factuality preference rankings and targets open-ended generation settings.

→ View original post on X — @dair_ai