RLAIF matches RLHF on summarization and dialogue tasks, with a direct-RLAIF variant achieving superior results by using LLM rewards directly during training.
Rlcd: Reinforcement learning from contrast distillation for language model alignment, 2023
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.CL 1years
2023 1verdicts
CONDITIONAL 1representative citing papers
citing papers explorer
-
RLAIF vs. RLHF: Scaling Reinforcement Learning from Human Feedback with AI Feedback
RLAIF matches RLHF on summarization and dialogue tasks, with a direct-RLAIF variant achieving superior results by using LLM rewards directly during training.