eval-fsr-a1-nemotron-bash-withtests-gpt5mini-swe-r431-rf0710-traces
收藏资源简介:
该数据集包含多轮对话记录,主要用于基于代理或模型的交互任务评估。数据特征包括:对话内容(conversations,含每轮的内容和角色)、代理标识(agent)、模型信息(model及model_provider)、日期(date)、任务类型(task)、集编号(episode)、运行ID(run_id)、试验名称(trial_name)、执行结果(result)、验证器输出(verifier_output)和追踪来源(trace_source)。数据集规模为2122个样本,仅包含训练集(train),总大小约285MB。适用于对话系统分析、代理行为评估或任务执行性能研究等场景。
This dataset contains multi-turn dialogue records, primarily used for evaluating agent- or model-based interactive tasks. The data features include: dialogue content (conversations, with each turns content and role), agent identifier (agent), model information (model and model_provider), date, task type, episode number, run ID, trial name, execution result, verifier output, and trace source. The dataset consists of 2122 samples, includes only a training set (train), with a total size of approximately 285MB. It is suitable for scenarios such as dialogue system analysis, agent behavior evaluation, or task execution performance research.




