eval-fsr-a1-stack-pytest-gpt5mini-swe-r467-rf0710-traces
收藏资源简介:
该数据集是一个结构化的多轮对话集合,专为评估和分析AI代理与模型的交互性能而设计。数据内容包含对话历史(conversations),其中每条记录由角色(role)和内容(content)组成,并关联了丰富的元数据:执行代理(agent)、使用的模型及提供商(model, model_provider)、对话日期(date)、任务类型(task)、实验序列标识(episode, run_id, trial_name)、任务结果(result)、验证器输出(verifier_output)以及数据来源追踪(trace_source)。数据集规模为3,681个训练样本,适用于对话系统评估、强化学习训练、AI行为分析或多模态任务性能研究等场景。
This dataset is a structured multi-turn conversation corpus, specifically designed for evaluating and analyzing the interaction performance between AI Agents and models. The data content includes conversation history (conversations), where each record comprises a role and content, and is linked to a rich set of metadata: executing agent (agent), the utilized model and its provider (model, model_provider), conversation date (date), task type (task), experiment sequence identifiers (episode, run_id, trial_name), task result (result), verifier output (verifier_output), and data source tracking (trace_source). The dataset contains 3,681 training samples, and is applicable to scenarios including dialogue system evaluation, reinforcement learning training, AI behavior analysis, and multimodal task performance research.
- 数据集名称:eval-fsr-a1-stack-pytest-gpt5mini-swe-r467-rf0710-traces
- 数据集地址:https://huggingface.co/datasets/laion/eval-fsr-a1-stack-pytest-gpt5mini-swe-r467-rf0710-traces
- 数据集大小:下载大小约 443 MB,数据集总大小约 553 MB
- 数据集分割:
- 训练集(train):包含 3,681 个样本
- 特征字段:
conversations:对话列表,每条对话包含content(字符串)和role(字符串)两个字段agent:字符串,智能体名称model:字符串,模型名称model_provider:字符串,模型提供方date:字符串,日期task:字符串,任务名称episode:字符串,回合编号run_id:字符串,运行IDtrial_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:默认配置名称为
default,数据文件路径为data/train-*




