DCAgent3/gaia_127_GLM_4_7_swesmith_sandboxes_with_tests_oracle_verified_120s_maxeps_131k5dec893b
收藏资源简介:
该数据集包含533个训练示例,主要特征包括对话记录(conversations,由角色和内容组成的列表)、代理(agent)、模型(model)、模型提供商(model_provider)、日期(date)、任务(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)和跟踪来源(trace_source)。数据集可能用于研究多轮对话系统、代理交互或任务执行评估,涉及不同模型和提供商的实验数据,每个示例记录了一次对话或任务的完整过程及其结果。数据以训练集形式提供,总大小约为25.1 MB。
This dataset consists of 533 training examples, with key features including conversations (a list comprising role and content), agent, model, model_provider, date, task, episode, run_id, trial_name, result, verifier_output, and trace_source. It is likely designed for research on multi-turn dialogue systems, agent interactions, or task execution evaluation, involving experimental data from various models and providers. Each example documents a complete dialogue or task process along with its outcome. The data is provided as a training split with a total size of approximately 25.1 MB.




