metadata
license: mit
datasets:
- PKU-Alignment/BeaverTails
language:
- en
base_model:
- PKU-Alignment/alpaca-7b-reproduced
This model is for the reproduction of results on Safe-RLHF dataset of paper "The crucial role of samplers in online direct preference optimization". Iteration 2 of DPO-mixp algorithm, trained on https://huggingface.co/zhezi12138/alpaca-7b-iter-1.