eval-fsr-a1-nemotron-rspec-swe-r406-rf0710-traces
收藏资源简介:
该数据集是一个多轮对话数据集,专门设计用于语言模型的训练、评估以及代理行为分析。每个数据样本代表一次完整的对话交互,其核心结构是一个对话列表(conversations),其中每条消息都包含内容(content)和角色(role)信息。此外,数据集还提供了丰富的元数据,包括使用的代理(agent)、模型名称(model)、模型提供方(model_provider)、对话日期(date)、任务类型(task)、运行标识符(如episode、run_id、trial_name)、任务结果(result)、验证器输出(verifier_output)以及数据来源追踪(trace_source)。数据集仅包含训练集部分,共有5385个样本,总大小约为823 MB。这些结构化的对话和元数据可用于深入分析模型在特定任务上的表现、对话流程的优化,以及多智能体交互的仿真研究。
This dataset is a multi-turn dialogue dataset specifically designed for language model training, evaluation, and agent behavior analysis. Each data sample represents a complete dialogue interaction, with its core structure being a conversation list (conversations) where each message includes content and role information. Additionally, the dataset provides rich metadata, including the agent used, model name, model provider, conversation date, task type, run identifiers (such as episode, run_id, trial_name), task result, verifier output, and data source trace (trace_source). The dataset contains only the training set, with a total of 5385 samples and a size of approximately 823 MB. This structured dialogue and metadata can be used for in-depth analysis of model performance on specific tasks, optimization of dialogue flows, and simulation research on multi-agent interactions.
- 数据集名称:laion/eval-fsr-a1-nemotron-rspec-swe-r406-rf0710-traces
- 数据集大小:下载大小约 675.45 MB,数据集总大小约 823.37 MB
- 数据集分割:仅包含训练集,共 5385 个样本
- 特征字段:
- conversations:对话列表,每条包含 content(字符串)和 role(字符串)
- agent:字符串,代理标识
- model:字符串,模型名称
- model_provider:字符串,模型提供方
- date:字符串,日期
- task:字符串,任务类型
- episode:字符串,回合标识
- run_id:字符串,运行 ID
- trial_name:字符串,试验名称
- result:字符串,结果
- verifier_output:字符串,验证器输出
- trace_source:字符串,追踪来源



