Alibaba-Apsara/Superior-Reasoning-SFT-gpt-oss-120b-Logprob

该数据集包含教师模型（gpt-oss-120b）为主要数据集Superior-Reasoning-SFT-gpt-oss-120b中的推理样本生成的令牌级对数概率。它是主数据集的配套数据集，通过唯一的sample_uuid进行关联。主数据集包含文本（提示、响应）、领域信息和高层次元数据，而本数据集（Logprobs数据集）则包含教师模型生成的令牌ID及其对应的对数概率值。数据格式结构化，包括示例UUID、令牌ID和对数概率值。该数据集在提升模型性能方面已证明有效，并提供了数据格式示例、许可证信息（CC BY 4.0）和引用文献。

This dataset contains the token-level log-probabilities generated by the teacher model (gpt-oss-120b) for the reasoning samples in the main Superior-Reasoning-SFT-gpt-oss-120b Dataset. It is a companion to the main dataset, with records linked via a unique sample_uuid. The main dataset contains the text (prompts, responses), domain info, and high-level metadata, while this Logprobs dataset contains the token IDs and their corresponding log-probability values from the teacher. The data follows a structured format, including sample UUID, token IDs, and logprobs values. The dataset has proven effectiveness in improving model performance, provides examples of data format, and is released under CC BY 4.0 license with citation information.