teamcore/DPO_Pm3B_RMAB_TG_clean_beta0.25dpo_proEurus_RM_7bbt_noise_flip_paper0.3_nu0.003

该数据集名为CR，包含了任务(task)、选择(chosen)、拒绝(rejected)等字段，以及对应的分布(chosen_distribution, rejected_distribution)和概率(bt_probs)。此外，还包括原始任务标记(is_original_task)、提示(prompt)、固定提示(fixed_prompt)、固定选择(fixed_chosen)、固定拒绝(fixed_rejected)和响应(response)。数据集分为default一个部分，共有2000个示例。The dataset named CR includes fields such as task, chosen, rejected, along with their respective distributions (chosen_distribution, rejected_distribution) and probabilities (bt_probs). It also includes original task markers (is_original_task), prompts (prompt), fixed prompts (fixed_prompt), fixed choices (fixed_chosen), fixed rejections (fixed_rejected), and responses (response). The dataset is split into one part, default, containing 2000 examples.