davidkling/hf-coding-tools-traces
收藏资源简介:
该数据集是HuggingFace AI Coding Tools的Agent Traces数据集,主要用于存储和展示AI编码工具的基准测试结果。数据集包含了32个会话,每个会话对应一个(工具、模型、努力程度、思考)配置,总计9,130个查询→响应轮次(约18,260个事件)。数据集涵盖了多种工具和模型,如claude_code、codex、copilot、cursor等工具,以及claude-opus-4-6、gpt-4.1等模型。数据格式采用Claude Code JSONL模式,每个会话包含用户和助手的交互记录,并附带基准测试的元数据。数据集来源于davidkling/hf-coding-tools-dashboard,并经过重新格式化以适应Hugging Face Agent Trace Viewer的使用需求。
This dataset rehydrates the benchmark results from davidkling/hf-coding-tools-dashboard into the JSONL session format consumed by the Hugging Face Agent Trace Viewer. It contains 32 sessions, one per (tool, model, effort, thinking) configuration, with a total of 9,130 query → response turns (≈18,260 events). The dataset covers tools like claude_code, codex, copilot, cursor, and models such as claude-opus-4-6, gpt-4.1, etc. Each session uses the Claude Code JSONL schema, including user and assistant interactions with benchmark metadata. The dataset is sourced from davidkling/hf-coding-tools-dashboard and reformatted for use with the Hugging Face Agent Trace Viewer.




