eval-laion_loopshape-30-8B_DCAgent_dev_set_v2-traces
收藏资源简介:
该数据集包含一系列AI代理或模型执行任务时产生的交互轨迹数据,以多轮对话形式组织。每条记录包括完整的对话历史(conversations字段,其中每轮对话有内容content和角色role),并记录了元信息如代理类型(agent)、模型名称(model)、模型提供商(model_provider)、任务类型(task)、执行日期(date)以及实验标识符(episode, run_id, trial_name)。此外,数据集包含任务执行结果(result)、验证器输出(verifier_output)和数据来源追踪信息(trace_source)。数据集规模为1272个训练样本,总大小约95MB,适用于AI代理行为分析、对话系统评估、任务完成轨迹研究等场景。
This dataset contains interaction trajectory data generated by AI agents or models while performing tasks, organized in multi-turn dialogue format. Each record includes complete dialogue history (conversations field, with each turn containing content and role), along with metadata such as agent type (agent), model name (model), model provider (model_provider), task type (task), execution date (date), and experiment identifiers (episode, run_id, trial_name). Additionally, the dataset includes task execution results (result), verifier output (verifier_output), and data source tracking information (trace_source). The dataset comprises 1272 training samples with a total size of approximately 95MB, suitable for scenarios like AI agent behavior analysis, dialogue system evaluation, and task completion trajectory research.
- 数据集名称: laion/eval-laion_loopshape-30-8B_DCAgent_dev_set_v2-traces
- 数据集来源: Hugging Face Datasets
- 数据集大小: 下载大小约 84.29 MB,数据集总大小约 95.21 MB
- 数据集分割: 仅包含训练集(train),共有 1272 个样本
- 数据特征:
conversations: 对话列表,每条对话包含content(字符串类型)和role(字符串类型)agent: 字符串类型,标识智能体model: 字符串类型,模型名称model_provider: 字符串类型,模型提供方date: 字符串类型,日期task: 字符串类型,任务名称episode: 字符串类型,回合标识run_id: 字符串类型,运行IDtrial_name: 字符串类型,试验名称result: 字符串类型,结果verifier_output: 字符串类型,验证器输出trace_source: 字符串类型,跟踪来源
- 配置文件: 默认配置(default),数据文件路径为
data/train-*




