scottgeng00/olmo-3-preference-mix-deltas_reasoning-chosen_qwen8b-yolo_scottmix-DECON

该数据集包含对话或文本选择的场景，每个样本由一个prompt（提示）和两个选择组成，分别为chosen（选中）和rejected（未选中）。每个选择包含content（内容）和role（角色）信息。此外，还记录了做出选择的模型(chosen_model和rejected_model)、数据集来源(dataset)和prompt的ID(prompt_id)，以及样本的分类(category)。训练集包含268,252个示例，数据集大小为5.88GB。The dataset consists of dialog or text selection scenarios, with each sample including a prompt and two options, chosen and rejected. Each option contains content and role information. Additionally, it records the model that made the selection (chosen_model and rejected_model), the source of the dataset (dataset), the ID of the prompt (prompt_id), and the category of the sample. The training set contains 268,252 examples, and the dataset size is 5.88GB.