Reinforcement Learning from AI Feedback (RLAIF)
More Factual LLMs: FactTune, a method to fine-tune LLMs for factual accuracy without human feedback
Large language models sometimes generate false statements. New work makes them more likely to produce factual output.
