InsultedByMathematics/rebel-ultrafeedback-test-evaluation-update-101

该数据集包含与提示和响应相关的多个特征，主要涉及多个响应、提示ID、提示内容、llama模型的提示及其对应的token序列。数据集还包含每个响应的奖励值、被选择的响应及其奖励值、被拒绝的响应及其奖励值，以及对应的llama模型输出和token序列。此外，数据集还包含微调后的响应以及被选择和被拒绝响应的对数概率。数据集的分割为test_prefs，包含1801个样本，总大小为106594758字节。

This dataset contains multiple features related to prompts and responses, including several responses, prompt IDs, prompt content, llama model prompts, and their corresponding token sequences. The dataset also includes reward values for each response, the chosen response and its reward value, the rejected response and its reward value, as well as the corresponding llama model outputs and token sequences. Additionally, the dataset contains fine-tuned responses and the log probabilities of the chosen and rejected responses. The dataset is split into test_prefs, containing 1801 samples with a total size of 106594758 bytes.