Samll/Llama-3.2-1B-Instruct_Taur_CoT_contrastive_relevances
收藏Hugging Face2025-02-14 更新2025-02-15 收录
下载链接：
https://hf-mirror.com/datasets/Samll/Llama-3.2-1B-Instruct_Taur_CoT_contrastive_relevances
下载链接
链接失效反馈官方服务：
资源简介：
这是一个包含多个配置的数据集，每个配置都是针对特定类型的推理任务，如演绎和归纳推理。数据集特征包括用户ID、正确答案、响应、响应标记、输出相关性、干扰项、干扰项相关性、原始cot响应、提示和响应以及cot起始位置等。此外，数据集还提供了零样本学习的数据拆分，用于训练模型进行无监督推理。
This is a dataset with multiple configurations, each tailored for specific types of reasoning tasks, such as deductive and inductive reasoning. Dataset features include user ID, correct answer, response, response tokens, output relevances, foil, foil relevances, original cot response, prompt and response, and cot start position. Additionally, the dataset provides a zero-shot data split for training models to perform unsupervised reasoning.
提供机构：
Samll
