eval-fsr-a1-stack-csharp-swe-r523-traces
收藏资源简介:
该数据集是一个包含多轮对话交互记录的数据集,主要用于记录和分析AI代理在特定任务中的表现。数据集包含3068个训练样本。每个样本的核心是一个对话序列(conversations字段),该序列由多个对话轮次组成,每个轮次包含content(对话内容)和role(发言者角色)信息。此外,每个样本还附带了丰富的元数据,用于描述对话发生的上下文和执行细节,包括:执行对话的agent(代理)、所使用的model(模型)及其model_provider(提供商)、对话发生的date(日期)、所执行的task(任务)、对话所属的episode(片段或情景)、run_id(运行标识符)、trial_name(试验名称)、task的result(结果)、verifier_output(验证器的输出结果)以及trace_source(数据跟踪来源)。数据集的结构表明它适用于对AI代理的对话能力、任务完成情况以及不同模型在不同任务场景下的表现进行分析、评估或基准测试。
This dataset is a collection of multi-turn dialogue interaction records, primarily used for recording and analyzing the performance of AI agents in specific tasks. It contains 3068 training samples. The core of each sample is a dialogue sequence (conversations field), consisting of multiple dialogue turns, each containing content (dialogue content) and role (speaker role) information. Additionally, each sample includes rich metadata to describe the context and execution details of the dialogue, including: the agent executing the dialogue, the model used and its model_provider, the date the dialogue occurred, the task performed, the episode to which the dialogue belongs, run_id, trial_name, the result of the task, verifier_output, and trace_source. The structure of the dataset indicates that it is suitable for analyzing, evaluating, or benchmarking the dialogue capabilities of AI agents, task completion, and the performance of different models across various task scenarios.
- 数据集名称:eval-fsr-a1-stack-csharp-swe-r523-traces
- 数据集地址:https://huggingface.co/datasets/laion/eval-fsr-a1-stack-csharp-swe-r523-traces
- 数据集提供方:LAION
- 数据集规模:下载大小为 375,401,343 字节,数据集总大小为 474,824,168 字节,包含 3,068 个训练样本。
- 数据划分:仅包含训练集(train)。
- 数据特征:
conversations:包含多个对话轮次,每个轮次有content(字符串类型)和role(字符串类型)字段。agent:字符串类型。model:字符串类型。model_provider:字符串类型。date:字符串类型。task:字符串类型。episode:字符串类型。run_id:字符串类型。trial_name:字符串类型。result:字符串类型。verifier_output:字符串类型。trace_source:字符串类型。
- 数据文件:训练数据存储在
data/train-*路径下。




