SteerLM: Attribute Conditioned SFT as an (User-Steerable) Alternative to RLHF.
Yi Dong, Zhilin Wang, Makesh Narsimhan Sreedhar, Xianchao Wu, Oleksii Kuchaiev
Browse the full EMNLP paper archive.
Yi Dong, Zhilin Wang, Makesh Narsimhan Sreedhar, Xianchao Wu, Oleksii Kuchaiev
Browse the full EMNLP paper archive.