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.
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>",
"messages": [
{"role": "system", "content": "You are a helpful AI assistant."},
{"role": "user", "content": "<task>"},
{"role": "assistant","content": "<full response with reasoning>", "reasoning_content": ""}
],
"input_tokens": <int>,
"output_tokens": <int>
}Generation settings
Distribution
Note: Math trajectories are longer on average (10.6K tokens) because the model produces detailed step‑by‑step reasoning. World Knowledge, Coding, and Agentic trajectories are shorter (~3.7–4K tokens). The max token ceiling of 16,384 was only approached in Math.
Files
Prompt Sources
License
Apache 2.0