five

laion/Qwen3-32B-SweSmith-traces

收藏
Hugging Face2026-03-12 更新2026-03-29 收录
下载链接:
https://hf-mirror.com/datasets/laion/Qwen3-32B-SweSmith-traces
下载链接
链接失效反馈
官方服务:
资源简介:
--- configs: - config_name: default data_files: - split: train path: data/shard-*.parquet --- # Qwen3-32B SweSmith RL Training Traces Training trajectories from RL training of Qwen3-32B on [SWEsmith](https://github.com/SWE-bench/SWEsmith) tasks. ## Details - **Model**: Qwen/Qwen3-32B - **Agent**: Terminus-2 - **Algorithm**: RLOO (rloo_n) - **Training steps**: 1-20 - **Total rows**: ~236K conversation episodes - **Tasks**: 2,500 SWEsmith software engineering tasks ## Columns | Column | Description | |--------|-------------| | `conversations` | List of `{role, content}` message dicts | | `agent` | Agent name (terminus-2) | | `model` | Model name | | `task` | Task identifier | | `episode` | Episode number within trial | | `trial_name` | Unique trial identifier | | `result` | Pass/fail result | | `instruction` | Task instruction text | | `verifier_output` | Test output from verifier |
提供机构:
laion
5,000+
优质数据集
54 个
任务类型
进入经典数据集
二维码
社区交流群

面向社区/商业的数据集话题

二维码
科研交流群

面向高校/科研机构的开源数据集话题

数据驱动未来

携手共赢发展

商业合作