eval-laion_lrboost-80-8B_DCAgent_dev_set_v2-traces
收藏资源简介:
该数据集记录了AI代理在多轮对话任务中的交互过程与执行结果。核心内容包括对话序列(conversations),每条消息包含内容(content)和角色(role)信息,并关联了执行代理(agent)、使用的模型(model)及其提供商(model_provider)、任务日期(date)、任务类型(task)、对话回合(episode)等元数据。此外,还记录了每次运行的唯一标识(run_id)、试验名称(trial_name)、任务执行结果(result)、验证器输出(verifier_output)以及数据追踪来源(trace_source)。数据集共包含1072个样本,仅提供训练分割(train),数据总量约为82.2MB。适用于分析AI代理在对话式任务中的行为模式、评估模型性能、研究多轮对话交互逻辑以及任务完成结果的验证与分析。
This dataset documents the interaction processes and execution outcomes of AI Agents during multi-turn dialogue tasks. Its core contents include conversation sequences (conversations). Each message contains content and role information, and is associated with metadata such as the executing agent, the utilized model (model), its provider (model_provider), task date (date), task type (task), and dialogue episode (episode). Additionally, it records the unique run identifier (run_id), trial name (trial_name), task execution result (result), verifier output (verifier_output), and data trace source (trace_source) for each execution. The dataset contains a total of 1072 samples, with only the training split (train) provided, and the total data volume is approximately 82.2 MB. It is applicable for analyzing the behavioral patterns of AI Agents in conversational tasks, evaluating model performance, studying the interaction logic of multi-turn dialogues, and verifying and analyzing task completion results.
- 数据集名称:eval-laion_lrboost-80-8B_DCAgent_dev_set_v2-traces
- 数据集地址:https://huggingface.co/datasets/laion/eval-laion_lrboost-80-8B_DCAgent_dev_set_v2-traces
- 数据集规模:
- 仅包含一个训练集(splits: train)
- 训练集样本数:1072 条
- 训练集大小:约 82.24 MB
- 下载大小:约 72.78 MB
- 数据特征:
conversations:对话列表,每条对话包含两个字段content:字符串类型,对话内容role:字符串类型,对话角色
agent:字符串类型,智能体名称model:字符串类型,使用的模型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,任务描述episode:字符串类型,轮次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,轨迹来源
- 数据格式:数据文件位于
data/train-*,默认配置名为default。




