Tongyi-ConvAI/EPO-RL-data

EPO-RL-data是用于训练战略推理模型Llama-3-8B-Instruct的数据集，通过强化学习（RL）方法进行训练。数据集中的dump.rdb文件用于在SOTOPIA平台上进行模型评估。EPO-RL-data is the dataset used for training the strategic reasoning model Llama-3-8B-Instruct, trained via Reinforcement Learning (RL). The dump.rdb file in the dataset is used for model evaluation on the SOTOPIA platform.