sleepyeldrazi/qwen3.6-27b-self-data-distillation-dataset

Qwen3.6-27B Self-Data-Distillation Trajectories Single‑turn reasoning trajectories generated by running Qwen3.6‑27B (via vLLM). Each trajectory contains a system prompt, a user task, and the model's full output (including reasoning steps embedded in the assistant content field). Data Format Four JSONL files, one per category. Each line is: { "id": "traj_<timestamp>_<idx>_<seq>", "source": "synthetic-qwen3.6-27b", "task": "<the prompt given to the model>"… See the full description on the dataset page: https://huggingface.co/datasets/sleepyeldrazi/qwen3.6-27b-self-data-distillation-dataset.