eval-fsr-a1-nemotron-csharp-swe-r352-rf0711-traces
收藏资源简介:
该数据集是一个专为人工智能代理(AI Agent)研究、评估和训练设计的多轮对话实验轨迹集合。其核心内容包括结构化的对话记录,每条记录包含一个由多轮次(`content` 和 `role`)构成的对话序列(`conversations`),并附带丰富的实验元数据,如执行对话的代理(`agent`)、使用的模型(`model`)及其提供商(`model_provider`)、实验任务类型(`task`)、实验日期(`date`)、实验轮次或情节(`episode`)、运行标识(`run_id`)、试验名称(`trial_name`)、任务执行结果(`result`)、验证器输出(`verifier_output`)以及轨迹来源(`trace_source`)。数据集共包含4521个样本,全部划分至训练集,适用于分析AI代理在不同任务下的对话行为、性能评估、轨迹学习或相关模型训练等场景。
This dataset is a collection of multi-turn dialogue experimental trajectories for AI Agent research, evaluation, or training. The core content consists of structured dialogue records, each containing a dialogue sequence (`conversations`) composed of multiple turns (`content` and `role`), along with rich experimental metadata. This metadata includes the agent executing the dialogue (`agent`), the model used (`model`) and its provider (`model_provider`), the type of experimental task (`task`), the experiment date (`date`), the episode or round of the experiment (`episode`), a run identifier (`run_id`), the trial name (`trial_name`), the task execution result (`result`), verifier output (`verifier_output`), and the trace source (`trace_source`). The dataset contains a total of 4521 samples, all allocated to the training set, making it suitable for analyzing AI agent dialogue behavior across different tasks, performance evaluation, trajectory learning, or related model training scenarios.
- 数据集名称:eval-fsr-a1-nemotron-csharp-swe-r352-rf0711-traces
- 发布机构:LAION(Large-scale Artificial Intelligence Open Network)
- 数据集大小:下载大小为 575,064,769 字节(约 548 MB),数据集总大小为 721,546,415 字节(约 688 MB)
- 数据划分:仅包含训练集(train),共 4,521 个样本
- 特征字段:
conversations:对话内容列表,每条对话包含content(字符串类型)和role(字符串类型)两个字段agent:代理名称(字符串类型)model:使用的模型名称(字符串类型)model_provider:模型提供方(字符串类型)date:日期(字符串类型)task:任务描述(字符串类型)episode:轮次(字符串类型)run_id:运行标识(字符串类型)trial_name:试验名称(字符串类型)result:结果(字符串类型)verifier_output:验证器输出(字符串类型)trace_source:追踪来源(字符串类型)
- 数据文件格式:数据存储在路径
data/train-*下,采用默认配置(default)




