XuHu6736/s1_54k_filter_with_isreasoning

Dataset Card for XuHu6736/s154kfilterwithisreasoning

Dataset Description

XuHu6736/s1_54k_filter_with_isreasoning is an enhanced version of the XuHu6736/s1_54k_filter dataset. This version includes additional annotations to assess the suitability of each question for reasoning training. These annotations, isreasoning_score and isreasoning, were generated using the deepseek-v3 model.

The purpose of these new fields is to allow users to filter, weight, or specifically analyze questions based on their evaluated reasoning potential.

For details on the original data sources, filtering, and composition of the parent dataset, please refer to the dataset cards for XuHu6736/s1_54k_filter and its predecessor XuHu6736/s1_59k.

Annotation Process

The additional annotations were generated as follows:

• —Model Used: deepseek-v3 (as specified by the dataset creator)
• —Input Field: The question field from the XuHu6736/s1_54k_filter dataset was used as {question_text}.
• —Prompt Used:

    Evaluate the suitability of the following question for reasoning training. Respond ONLY with an integer score between 0 and 100. Do not include any other text, explanation, or labels.

    Question: {question_text}

• —Output Fields:
• —isreasoning_score: An integer score between 0 and 100, directly from the model's response, indicating the question's suitability for reasoning training. A higher score suggests greater suitability.
• —isreasoning: A label derived from the isreasoning_score (e.g., potentially a boolean or categorical classification like "High", "Medium", "Low" based on score thresholds defined by the dataset creator). Users should verify how this label is specifically derived if using it for critical tasks.