DCAgent/e1_gpt_long_swegym_20k_glm47_traces

该数据集包含20,000个训练示例，主要用于对话或任务执行场景。每个示例包含对话记录（conversations，含内容和角色）、代理信息（agent）、模型详情（model和model_provider）、日期（date）、任务类型（task）、剧集（episode）、运行ID（run_id）、试验名称（trial_name）、结果（result）、验证器输出（verifier_output，当前为空）、跟踪来源（trace_source）和SWEGYM来源（swegym_source）。数据集可能涉及多轮交互或自动化任务评估。

This dataset consists of 20,000 training examples, primarily designed for dialogue or task execution scenarios. Each example includes conversation records (conversations, with content and role), agent information (agent), model details (model and model_provider), date (date), task type (task), episode (episode), run ID (run_id), trial name (trial_name), result (result), verifier output (verifier_output, currently null), trace source (trace_source), and SWEGYM source (swegym_source). The dataset likely involves multi-turn interactions or automated task evaluation.