poolside-laguna-hackathon/py-bug-trace-laguna-xs-2-l3-rollouts
收藏资源简介:
这是一个用于评估AI模型在代码生成或任务完成场景下表现的数据集。数据集包含多个特征字段,如trace_id(跟踪ID)、example_id(示例ID)、rollout_number(展开编号)、task(任务)、prompt(提示,包含角色和内容)、completion(完成,包含角色、内容和推理内容)、answer(答案)、reward(奖励)、score(分数)、correct(正确性)等。其中,info字段提供了任务的元信息,包括难度级别、类别、任务库版本和测试代码,而pytest_reward字段表明与Python测试框架相关。数据集可能用于训练或测试模型在编程任务中的性能,通过奖励和分数来衡量模型输出质量。数据分片为训练集,包含15个示例,总大小约97KB。
This is a dataset designed for evaluating AI model performance in code generation or task completion scenarios. It includes multiple features such as trace_id, example_id, rollout_number, task, prompt (with role and content), completion (with role, content, and reasoning content), answer, reward, score, and correct. The info field provides metadata about tasks, including difficulty level, category, task bank version, and test code, while pytest_reward indicates relevance to the Python testing framework. The dataset is likely used for training or testing models on programming tasks, with rewards and scores measuring output quality. It is split into a training set with 15 examples and a total size of approximately 97KB.




