Towards Pareto-Efficient RLHF: Paying Attention to a Few High-Reward Samples with Reward Dropout.

Changhun Lee, Chiehyeon Lim

Browse the full EMNLP paper archive.