eval-fsr-a1-swegym-swe-r380-rf0710-traces
收藏资源简介:
该数据集记录了AI模型或智能体在特定任务环境中的交互对话过程。数据集包含2,264个训练样本,总大小约为622MB。每个样本包含完整的对话历史(conversations字段,记录每条消息的内容和发送者角色)、参与交互的智能体信息(agent字段)、使用的模型及其提供商(model和model_provider字段)、实验执行时间(date字段)、任务类型(task字段)、实验轮次标识(episode, run_id, trial_name字段)以及任务执行结果(result字段)。此外,还包含验证器输出(verifier_output字段)和轨迹来源信息(trace_source字段)。该数据集适用于分析AI模型在任务导向对话中的行为模式、评估模型性能、研究多轮对话交互,或作为训练对话系统的数据源。
This dataset records the interactive dialogue processes of AI models or agents in specific task environments. It contains 2,264 training samples with a total size of approximately 622MB. Each sample includes complete dialogue history (conversations field, recording the content and sender role of each message), information on the participating agents (agent field), the model used and its provider (model and model_provider fields), experiment execution time (date field), task type (task field), experiment round identifiers (episode, run_id, trial_name fields), and task execution results (result field). Additionally, it includes verifier output (verifier_output field) and trace source information (trace_source field). This dataset is suitable for analyzing behavioral patterns of AI models in task-oriented dialogues, evaluating model performance, studying multi-turn dialogue interactions, or serving as a data source for training dialogue systems.
- 数据集名称:laion/eval-fsr-a1-swegym-swe-r380-rf0710-traces
- 数据集大小:下载大小为 314,884,265 字节(约300MB),数据集大小为 621,694,319 字节(约593MB)
- 数据划分:仅包含一个训练集(train),共 2,264 个样本
- 特征字段:
- conversations:对话内容列表,包含 content(字符串)和 role(字符串)
- agent:字符串
- model:字符串
- model_provider:字符串
- date:字符串
- task:字符串
- episode:字符串
- run_id:字符串
- trial_name:字符串
- result:字符串
- verifier_output:字符串
- trace_source:字符串
- 数据文件路径:data/train-*(位于根目录下)
- 配置文件:仅有一个配置名称为 default




