eval-laion_swesmith-coldstart-complete-lt32k-2ep-8B_DCAgent_dev_set_v2-traces
收藏资源简介:
该数据集是一个结构化记录集,主要用于捕获和存储AI代理在特定任务上的多轮交互对话及其执行元数据。数据集核心包含多轮对话记录,每轮记录包括对话内容和角色信息,并附带丰富的实验元数据,如使用的代理类型、模型信息、模型提供商、执行日期、任务类型、场景片段、运行ID、试验名称、执行结果、验证器输出以及数据来源追踪。数据集规模为1511个样本,存储为训练集分割,总数据量约130MB。它适用于分析AI代理的对话行为、任务执行性能评估、多轮交互模式研究以及实验结果的追溯与验证。
This dataset is a structured record collection primarily used for capturing and storing multi-turn interactive dialogues of AI agents on specific tasks along with their execution metadata. The core of the dataset includes multi-turn dialogue records, each containing content and role information, and is accompanied by rich experimental metadata, such as the agent type used, model information, model provider, execution date, task type, scenario segment, run ID, trial name, execution result, validator output, and data source tracking. The dataset scale is 1511 samples, stored as a training set split, with a total data volume of approximately 130MB. It is suitable for analyzing AI agent dialogue behaviors, evaluating task execution performance, researching multi-turn interaction patterns, and tracing and validating experimental results.
- 数据集名称:
eval-laion_swesmith-coldstart-complete-lt32k-2ep-8B_DCAgent_dev_set_v2-traces - 来源地址: https://huggingface.co/datasets/laion/eval-laion_swesmith-coldstart-complete-lt32k-2ep-8B_DCAgent_dev_set_v2-traces
- 数据集大小: 下载大小为 116,648,294 字节,数据集总大小为 130,361,882 字节
- 数据划分: 仅包含训练集(train),共 1,511 个样本
- 数据特征:
conversations: 列表类型,包含content(字符串)和role(字符串)字段agent: 字符串,代理标识model: 字符串,模型名称model_provider: 字符串,模型提供者date: 字符串,日期task: 字符串,任务类型episode: 字符串,回合编号run_id: 字符串,运行标识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
- 配置文件: 默认配置(default),数据文件路径为
data/train-*




